Professional Documents
Culture Documents
NREL/TP-500-41160
Simulation of a Baseline 750-kW February 2009
Wind Turbine Using State-of-the-
Art Simulation Codes
F. Oyague
Gearbox Modeling and Load Technical Report
NREL/TP-500-41160
Simulation of a Baseline 750-kW February 2009
Wind Turbine Using State-of-the-
Art Simulation Codes
F. Oyague
Prepared under Task No. WER8.2001
This report was prepared as an account of work sponsored by an agency of the United States government.
Neither the United States government nor any agency thereof, nor any of their employees, makes any
warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or
usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not
infringe privately owned rights. Reference herein to any specific commercial product, process, or service by
trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement,
recommendation, or favoring by the United States government or any agency thereof. The views and
opinions of authors expressed herein do not necessarily state or reflect those of the United States
government or any agency thereof.
Printed on paper containing at least 50% wastepaper, including 20% postconsumer waste
Executive Summary
The wind energy industry continually evolves, and industry professionals have streamlined
gearbox design to a consensus configuration. This configuration and its design iteration have
existed for many years; consequently, design and manufacturing flaws have been minimized
sequentially. Regardless of the maturity of the gearbox design and design process, however, most
wind turbine downtime is attributed to gearbox-related issues. Moreover, gearbox replacement
and lubrication accounts for 38% of the parts cost of the entire turbine.
Several hypotheses have been offered to explain gearbox failure, including the absence of a
number of load cases relevant to the design process; the transfer of nontorsional loads between
the different components of the drivetrain; the lack of a uniform standardization of bearing-life
analysis calculations; and poor communication between wind turbine designers, gearbox
suppliers, and bearing providers.
This report discusses determining a method for revealing the missing loading conditions that
should be factored into the gearbox-design process. This objective is achieved by development of
a number of analytical models that sequentially increase in complexity, and which are capable of
reproducing the dynamical behavior of the internal components of the drivetrain. Additionally,
the parameters obtained from these models are correlated with the gearbox-design process.
Importantly, the models developed are offered freely to improve communication and to open
information-sharing avenues between manufacturers and designers involved in the non-vertical
design process.
The models reveal that the level of complexity does not greatly affect torsional behavior.
Furthermore, models of higher complexity are capable of providing important insight into the
loading conditions for the bearings of the gearbox, and still account for loads generated by gear-
tooth interactions.
i
Acknowledgements
I would like to thank my primary supervisor Professor Dr. Dipl.-Ing. Martin Khn, my secondary
supervisor Dipl.-Ing. Stefan Hauptmann, and my COMMAS supervisor Professor Dr.-Ing. Bernd
Krplin, for arranging the integration of the COMMAS Masters Program with the Endowed
Chair of Wind Energy (SWE), and thus allowing me the opportunity to participate in cutting-
edge research.
Thanks also go to Chief Engineer Sandy Butterfield, my external supervisor, who allowed me the
opportunity to research and write this report at the National Renewable Energy Laboratory
(NREL), and who provided constant advice and guidance; and to Ed Hahlbeck, Don McVittie,
and Brian McNiff, who shared their expertise in the gearbox-design process and the wind turbine
industry, and gave me valuable advice.
I thank INTEC GmbH for providing the multibody system code SIMPACK, which made
possible the development of the progressive models used in this report.
Last, but not least, I thank my family and friends for their ever-constant support and guidance.
Specifically, I would like to acknowledge Maria Christina and Marnix Vanderplas, who
generously provided the financial support that enabled me to produce this report.
ii
Table of Contents
Symbols ......................................................................................................................................... vi
Introduction................................................................................................................................... 1
Motivation................................................................................................................................... 1
Problem Definition...................................................................................................................... 2
Approach..................................................................................................................................... 3
Overview..................................................................................................................................... 4
Couplings.............................................................................................................................. 10
Gearbox ................................................................................................................................ 10
Planetary Gearbox............................................................................................................. 11
Brakes ................................................................................................................................... 13
Aerodynamic Brakes......................................................................................................... 13
Mechanical Brakes............................................................................................................ 14
Generator .............................................................................................................................. 14
Pitch Control......................................................................................................................... 15
Stall Control.......................................................................................................................... 15
Gears ............................................................................................................................................ 16
Wear ..................................................................................................................................... 19
Abrasion ............................................................................................................................... 20
Surface Fatigue..................................................................................................................... 21
Micropitting.......................................................................................................................... 21
Macropitting ......................................................................................................................... 22
Spalling................................................................................................................................. 22
Crushing ............................................................................................................................... 23
iii
Plastic Flow .......................................................................................................................... 23
Fracture................................................................................................................................. 23
Bearings .................................................................................................................................... 24
Logarithmic Descent................................................................................................................. 34
Blade Inertia.............................................................................................................................. 37
Shaft Stiffness....................................................................................................................... 49
Inertias .................................................................................................................................. 50
Shaft Stiffness....................................................................................................................... 54
Inertias .................................................................................................................................. 54
Stiffness..................................................................................................................................... 55
iv
FAST Model Description............................................................................................................ 58
Model Comparison.................................................................................................................... 61
Vibration Analysis................................................................................................................ 68
Controls ................................................................................................................................ 68
Drivetrain Specification........................................................................................................ 69
Design Selection................................................................................................................... 69
Future Work................................................................................................................................ 74
References.................................................................................................................................... 76
Appendix:..................................................................................................................................... 78
Aerodynamic Simulation............................................................................................................ 78
v
Symbols
b With of cross-sectional area [m]
c Airfoil cord length [m]
c Damping coefficient [N/m/sec]
Cl Lift coefficient [-]
Cd Drag coefficient [-]
D Drag force [N]
dA infinitesimal element [-]
Dring Ring gear diameter [m]
DSun Sun gear diameter [m]
E Youngs modulus [N/m2]
G Shear modulus of elasticity [N/m2]
h Height of the cross-sectional [m]
Ia Area moment inertia [m4]
Im, J Mass moment of inertia [kg m2]
Ip Polar moment of inertia [m4]
k Spring constant [Nm/rad]
K Stiffness [Nm/rad]
keq Equivalent stiffness [Nm/rad]
kt Torsional stiffness [Nm/rad]
l Length of rod under torsion [m]
L Lift force [N]
ls Airfoil span [m]
m Mass [kg]
m Mass flow rate [kg/sec]
ma Mechanical advantage [-]
mG Gear ratio [-]
mv Velocity ratio [-]
nHss High-speed shaft angular velocity [-]
nLss Low-speed shaft angular velocity [-]
P Period of oscillation [sec]
in Angular velocity input [rad/sec]
out Angular velocity output [rad/sec]
rin Pitch radius of input gear [m]
rout Pitch radius of output gear [m]
r1 Radius of rod under torsion [m]
ymax Maximum beam deflection [m]
Trot Rotor thrust [N]
V1 Incoming wind velocity [m/s]
V3 Outgoing wind velocity [m/s]
U Undisturbed fluid velocity [m/sec]
Density [kg/m3]
Stress [N/m2]
Strain [-]
Shear strain deformation [rad]
vi
Shear stress [N/m2]
Angle of torsional deflection [rad]
T Applied torque [N/m]
n Undamped natural frequency [rad/sec]
n Undamped natural frequency [HZ]
x Acceleration [m/sec2]
x Velocity [m/sec]
x Position [m]
Damping ratio [-]
Logarithmic descent [-]
Rotation around the X axis [-]
vii
List of Figures
Figure 1. Downtime hours accumulated from 2003 to 2007 for wind turbines operating in
Germany ................................................................................................................................. 1
Figure 33. Equivalent one-stage drivetrain with single stiffness and inertia................................ 39
Figure 39. True response (left) and collected response (right) ..................................................... 45
Figure 41. Stage 1 simulated response with refined integration time step ................................... 46
viii
Figure 44. Subsequent gearbox stages topology........................................................................... 49
Figure 45. Stage 2 model simulated response with fine integration time step ............................. 51
Figure 49. Stage 3 model simulated response with fine integration time step ............................. 55
Figure 51. Planetary stage with added degrees of freedoms and force constraints ...................... 57
Figure 57. The input torque and output torque for the Stage 1 model.......................................... 62
Figure 58. Input torque and output torque comparison for Stage 3 .............................................. 63
ix
Introduction
The growing consciousness of global climate change has helped the importance of renewable
energy become paramount. The wind industry, in particular, has seen an unquenchable demand
as a result, and the need for reliable and affordable wind turbines now is all the more apparent.
Unfortunately, recurrent drivetrain failures have characterized the industry and have prevented
the turbines from achieving their intended 20-year design life (see Figure 1).
65000
60000
55000
50000
45000
40000
Stop Hours
35000
30000
25000
20000
15000
10000
5000
0
Rotor
Air Brake
Mech. Brake
Pitch Adjust.
Main shaft/bearing
Gearbox
Generator
Yaw System
Windvane/anemometer
Elec. Controls
Elec. System
Hydraulics
Sensors
The component most responsible for downtime is the gearbox [1]. Gearbox replacement and
lubrication account for 38% of the parts cost for the entire turbine system [2]. This situation calls
for the implementation of new and advanced simulation techniques to be integrated into the
gearbox-design process so that this component can meet its intended design life.
Motivation
This report is a combination of the efforts of the National Renewable Energy Laboratory (NREL)
(United States) and the Endowed Chair of Wind Energy (SWE) at the University of Stuttgart
(Germany). It is part of the gearbox reliability collaborative (GRC) project, which seeks to
improve gearbox reliability and, consequently, wind turbine reliability. The GRC originates from
NRELs National Wind Technology Center (NWTC). The GRC attempts to bring together the
various parties involved in the gearbox design process with the common goal of the improve
ment of the lifetime of gearboxes. It seeks to achieve this by exploring three avenues of research:
drivetrain numerical analysis and modeling, full-scale dynamometer testing, and field testing.
Problem Definition
The drivetrain has the important task of transforming the rotational energy of the rotor into
electrical power. The drivetrain is composed of several elements, each of which contributes to a
specific task. Except for the direct drivetrain, all drivetrains have a gearbox. The gearbox is
responsible for increasing the angular velocity transmitted from the rotor to the generator, to
satisfy the velocity required by the generator. It is the component of greatest mechanical
complexity in the drivetrain and, as noted, is responsible for most wind turbine operational
downtime and for increased costs.
Several hypotheses have been offered to explain the early failure of gearboxes. Among these is
the possibility that a number of load cases are not considered in the loads document used for the
design of the independent components of the drivetrain. There also is the possibility that
nontorsional loads and dynamical effects are transferred among the components of the drivetrain.
Lastly, it is possible that the reliability of bearings is not uniform throughout the industry. This is
due to an incomplete definition of the assumptions and parameters specified in the current
bearing standards that subsequently, has led individual manufactures to create in-house codes for
predicting bearing life.
Another important characteristic of gearbox failure is that it occurs in turbines of several sizes
having the same or similar configurations. It is believed that problems present in previous turbine
models (ranging from 500 kW to 1000 kW) have persisted, and therefore they still occur in
todays larger turbines. This is very relevant, because if the problem in smaller turbines was
solved then the solution could be extrapolated to larger turbines. This situation is advantageous
in that working with smaller turbines reduces both the cost of use and the availability of test
subjects presently operating in the field. Furthermore, this translational property between
turbines enabled the use of experimental data collected from a 750-kW turbine in the field to
validate the progressive models described in this report.
Although the presence of problems involving the gearbox is evident, the characterization and
reasons for these problems are not as clear. Several observations have been made with respect to
gearbox failures in an attempt to better characterize the problem. For example, over the years
gearboxes have been streamlined by several independent manufacturers and progressively have
converged to a similar configuration. This configuration thus represents a mature consensus
design of manufacturers, and it is used almost universally in the industry. Failures occur in
gearboxes of this configuration independent of the manufacturer and any slight differences in the
actual design. This suggests that each independent manufacturer is performing the same design
routine. Additionally, because the industry is highly competitive and because it can be assumed
that manufacturers are capable of correcting quality-control problems, it follows logically that
the failures experienced are not caused by poor workmanship. Therefore the problems inherent to
gearboxes appear to be a product of flaws in the design process.
parties involved in the process must have a complete knowledge of the behavior of all
components. This necessitates, for example, that the wind turbine manufacturers understand the
behavior of the gearbox, that the gearbox manufactures understand the loading conditions and
interactions of the drivetrain and the gearbox, and that the bearing manufactures understand the
interaction of the internal components of the gearbox. Essentially, a level of absolute
transparency and comprehensive information sharing is required for the design to succeed.
Unfortunately, sharing knowledge is very difficult due to proprietary information barriers; there
is competition between manufacturers and suppliers because they share common interests and
clients. A single wind turbine company might have more than one gearbox supplier, for example,
or a single gearbox supplier might have more than one bearing supplier. Conversely, the gearbox
supplier might be working with multiple wind turbine manufacturers. It is this level of intense
competition that makes transparency and information sharing virtually impossible.
Approach
This report covers a small portion of the enormously detailed analytical task of the gearbox
reliability collaborative. Its approach seeks to integrate into the drivetrain-design process several
numerical models which capture the dynamical nature of the drivetrain. These dynamic models
progressively build in complexity, and aim to provide insight into the internal forces inherent to
the dynamical behavior of the drivetrain. Additionally, the progressive nature of these models
offers a method of filtration for sensitive information shared among the different parties of the
design processultimately increasing the transparency of the design process.
Due to the reciprocal characteristics of the interactions among the dynamical components of the
wind turbine, a fully coupled model which integrates all of these components has proven
valuable [32]. The scope of this report is limited to the drivetrainmore specifically to the
gearbox. The limited scope required the integration of different simulation codes to more closely
mimic the behavior of a fully coupled model. These codes included aeroelastic codes used to
create loading conditions for the gearbox models (described below).
The implementation of aeroelastic software has been extremely popular due to its ability to
capture the interactions between inertial, elastic, and aerodynamic forces. This type of simulation
can predict the general behavior of the wind turbine and its interactions with the wind. In this
type of simulation the drivetrain typically is extremely simple. It provides a good approximation
for the overall behavior of the entire turbine, although it neglects the interaction of the
drivetrains internal components. Due to the proven capabilities of aeroelastic software to closely
represent wind turbine systems, this software was used in combination with a multibody system
software.
The multibody system simulation approach was chosen to generate the models due to its ability
to represent the drivetrain in a more detailed manner. This level of detail reveals important
information of the internal loading conditions resulting from the dynamical interactions. This
approach represents each component of the drivetrain as a rigid body, and defines its interaction
with the other components. The representation of each component as a rigid body reduces the
number of degrees of freedom (DOF), which enables the simulation to be fast and
computationally light.
The integration of these two powerful tools resulted in progressive comprehensive models,
capable of generating load cases for the individual components of the drivetrain. Ideally, these
load cases should be integrated into the standard gearbox-design process, because typically the
process does not fully account for the dynamical behavior of the individual components of the
drivetrain. Additionally, the models were compared to each otherfrom least complex to most
complexto identify the level of detail and fidelity that each model could generate. This
information can be used by a gearbox designer to select the drivetrain model complexity that is
most appropriate for an individual task.
Two primary structural classifications of wind turbines commonly are used: the vertical axis
wind turbine and horizontal axis wind turbine. The advantage of vertical axis turbines (Figure 2)
is that they are omni-directionalthey have the ability to accept wind from every direction. This
eliminates the problem of orienting the rotor with respect to the wind. Vertical axis turbines also
have drivetrains that do not meet their expected design life. Independent of the advantages of the
vertical axis configuration, horizontal axis turbines (Figure 3) have proven to be more efficient.
Not surprisingly the horizontal axis design currently is the most popular design in the industry
and is the design upon which this report focuses. [25]
Modular Drivetrain
Currently, most operating turbines follow the modular configuration. All individual components
of the drivetrain are mounted onto the bedplate, and the bedplate is designed to be torsionally
stiff. Nevertheless, there is debate surrounding the bedplates actual behavior that suggests that it
is not as stiff as it should be, and that its flexibilities influence not only the interaction between
the different components of the drivetrain but also its vibrational behavior [28]. The main
components of the drivetrain are the rotor shaft or low-speed shaft, the gearbox, the brakes, and
the generator.
The modular configuration allows a non-vertical design process, which means that different
suppliers can contribute to the development of the different components of the drivetrain. This
inherently reduces the overall cost by fomenting a competitive environment among suppliers,
and reduces the in-house requirements of the turbine manufacturer. Figure 4 shows the typical
configuration for a modular drivetrain [3].
High-Speed Generator
Shaft
Brake
Gearbox
Main Main
Hub Bearing Shaft
Bedplate
Integrated Drivetrain
As its name indicates, the integrated drivetrain includes the components of the modular type
discussed above. Generally, the gearbox becomes the main component of the system and the
remaining components are attached or flanged to it. This compact design is not dependent on the
bedplate, thus making the drivetrain lighter and preventing misalignment of the shafts and other
components.
The main disadvantage of this design is that a defective part generally results in the dismount of
the entire nacelle, making its maintenance extremely expensive. Additionally, the gearbox
becomes a special and critical component of the system with the capability of holding the rotor,
therefore the housing construction must be robust. In this design it is difficult to entirely isolate
the reaction forces from the rotor, therefore some of these forces could be transferred onto the
other components of the drivetrain, thus reducing their operating lives.[3]
With respect to the design process, the integrated drivetrain configuration generally is limited to
one gearbox design. The design process therefore must follow a vertical structure with the wind
turbine manufacturers being closely involved throughout the entire design process. To a certain
degree this prevents cooperation with different gearbox providers and generally results in the
turbine manufacturers owning the gearbox designs; this can result in increased cost. Figure 5
shows the typical configuration for the integrated drivetrain.
Brake
Generator
Hub
Gearbox
Brake Generator
Bed Plate
In the direct drivetrain approach the turbine manufacturer designs the entire drivetrain, resulting
in a vertical design approach. The only interaction generally is between the turbine manufacturer
and the generator provider, and involves discussion regarding size constraints and coupling
requirements. Although this configuration eliminates the prominent gearbox problems, it
nevertheless seems to be an expensive solution. Consequently, the industry for the most part has
maintained the construction of the modular configuration. This concept, however, has been
implemented by some European companies such as Enercon (see Figure 7 for a direct drive
example).
Main carrier
Blade flange
Generator
Pitch drive
Tower
The low-speed shaft supports the weight of the rotor and transmits all the reaction forces to the
main frame through the main bearing. These reaction forces are composed of all nontorsional
loads such as axial thrust from the wind, as well as disturbances from turbulence caused by
uneven wind distribution or wind shear.
While operating, dynamical effects such as vibration imbalances from the blades and gyroscopic
loads from yaw movement also contribute to nontorsional loading. Although the low-speed shaft
is a simple mechanical device it is very important, because the unintended transmission of
reaction forces to the other components of the drivetrain could reduce operating life [5].
The simplest model of the low-speed shaft is a totally rigid body that allows no deflections. A
better approximation of its real behavior would include its torsional deflection, although it still
would be modeled as being infinitely rigid to bending. This approach provides a worse-case
scenario for the load transmission to the rest of the component of the drivetrain, because
nontorsional loads would be transferred directly with no absorption from the low-speed shaft. A
more comprehensive approach includes flexible representation of the low-speed shaft; this truly
would reveal the contribution of the low-speed shaft to the reduction of the transmission of
nontorsional loads to the rest of the drivetrain.
Couplings
Couplings are mechanical devices used to connect shafts together, and they are capable of
transmitting a torsional load from one shaft to the other. The two main types of couplings are
rigid coupling and compliant or flexible coupling.
A rigid coupling locks both shafts together, thereby allowing no relative motion between the two
shafts. The disadvantage of this type of coupling is that the shaft axis must be precisely aligned
to prevent the transmission of forces and moments other than torsion.
The compliant or flexible coupling relies on the use of elastomers or other materials to dampen
transmitted torsional loads without shock-load excursions. Another advantage is that to a certain
degree they can handle axial, angular, parallel, and torsional misalignment. Thus the assembling
tolerances of the system are not required to be as precise. Its dampening capabilities also help
decrease torque spikes and vibrations that can damage other components of the drivetrain. The
main disadvantage of this coupling is that it can increase the level of backlash in the drivetrain or
torsional clearance [6].
Gearbox
The gearbox is a mechanical device capable of transferring torque loads from a primary mover to
a rotary output, typically with a different relation of angular velocity and torque. In the case of
wind turbines the gearbox connects the low-speed shaft and the generator; therefore its gear ratio
generally is dictated by the requirement of the generator and the angular velocity of the rotor.
In the case of electrical power production with an asynchronous generator, the output of the
gearbox (which is connected to the generator) usually operates in the ranges from 50Hz to 60 Hz
or 1,500 rpm to 1,800 rpm. This depends on the frequency of the grid to which the generator is
connected, and on the number of poles of the generator [5]. In many of todays modern machines
the generator is able to operate at a greater range of speeds. Regardless of this advantage a speed
increment still is required, although it might not be as significant as that needed for the
asynchronous generator.
The angular velocity of the rotor can be linked to the tip speed ratio. The tip speed ratio is the
ratio of the velocities of the tip of the blade to the wind velocity. An important factor that
dictates the tip speed ratio is the area covered by the blades with respect to the area swept by the
rotor. As the area of the blades is reduced, the tip speed ratio must be increased. The tip speed is
directly proportional to the radius of the rotor, therefore this dimension also dictates rotor
angular velocity [3].
Over the years, the power output capacity of turbines has been improved by increasing both the
slenderness of the blades and the rotor sweep area. The angular velocities therefore are reduced,
creating the need for gearboxes that are capable of handling greater torques and greater gear
ratios. Greater torque has led to the development of larger gearboxeswhich also have been
characterized by the same failures as those experienced by smaller configurations used in smaller
machines. Moreover, due to the larger size of the components larger gearboxes also might
10
experience material properties scaling discrepancies. Greater gear ratios have caused the use of
larger gear ratios per stage and an increase in the number of stageswhich has created more
complex configurations that are prone to failures.
There are two main types of gearboxes, parallel shaft gearboxes and planetary gearboxes. These
are described in the following sections.
In parallel gear stages, the gear interaction can be modeled in several ways (explained below).
The parallel stages also have an advantage in that they can be modeled with a minimum number
of degrees of freedom. This is because they are attached directly to the gearbox housing and in
many cases are attached to the inertial frame. This results in the calculation of the angular
velocities of the components of the stage with respect to a fixed frame, thus easing its
comparison and understanding.
Planetary Gearbox
The epicyclic or planetary gearbox offers several advantages compared to the parallel-shaft
configuration, including a higher gear ratio in a smaller package. This compact configuration
has the advantage of reducing the overall mass of the gearbox, which is an important requirement
for wind turbine gearboxes, because the head mass of the turbine is kept to a minimum.
Additionally, the planetary configuration has the capability of handling greater torque loads. This
is because the load is distributed or shared by the number of planet gears, therefore more teeth
always are in contact. Another advantage is its geometrical configuration. The input and output
have a concentric axis and the same rotational direction; therefore, it is very simple to build
multistage planetary boxes that maintain a streamlined and compact design.
The planetary box is more complicated than the parallel shaft, because it is composed of three
moving components per stage. These components include the planet gear, the planet carrier, and
the sun pinion. The ring gear is also part of the planetary box however it is fixed to the gearbox
housing.
The left side of Figure 8 shows the anterior view of a planetary stage. In the outer section the
ring gear is represented by circular mesh. Generally the ring gear is fixed to the housing and it is
not a moving component. It is an internal gear; the teeth are on the inside of the ring and mesh
with the teeth of the planets at all times. The planet carrier is rendered in blue and holds four
planets (shown in darker gray). The planet carrier ensures that the proper ring-planet and planet-
sun center distances are maintained. Its configuration generally is more robust than that shown in
the figure, because it must undertake the high-input torque loads.
11
The planets are supported on the planet carrier by shafts with bearings, therefore the planets can
rotate freely with respect to the planet carrier. The input is provided to the planet carrier which
distributes the torsional load among the planets so that the ring-sun interaction can be completed.
The posterior side of the planetary stage is shown on the right side of Figure 8. The sun is located
in the center of the planets. The load from the planets is transferred to the sun at four points to
become the output of the stage.
Planets
Models that depict the planetary gearboxes typically increase in complexity with respect to the
parallel shaft configuration; this is due to the interaction of the different moving components and
reference frames. A simple example is the behavior of the planets interaction with the planet
carrier; in a simple model the planets are assumed to have only one degree of freedom (rotation).
This relation shows the planets behavior with respect to the planet carrier. At the same time, the
planets are interacting with two other components the ring gear (which is fixed to the inertial
frame) and the sun (which is constrained to the reference frame but allowed to rotate). This
setting leads to difficult data interpretation because the contribution of all the acting components
must be considered to be able to understand the behavior. Simple torsional models present their
own difficulties, because the planetary stage must be simplified and specific relations among
components must be created to simulate the desired response.
As noted, the rotational direction of the input and output of a single planetary stage are the same.
This can be seen clearly in Figure 9. For the configuration below, the gear ratio is given by the
following expression.
n Hss DRing
= 1+ (1)
n Lss DSun
Where nHss is the rotational speed of the high-speed shaft of output, nLss is the rotational speed of
the low-speed shaft or input, Dring is the pitch diameter of the ring or its number of teeth, and
Dsun is the pitch diameter of the sun or its number of teeth [5].
12
Ring gear
Planet carrier
rotation
Notice that, in the expression used to calculate the gear ratio, the only parameter taken into
account is the dimensions of the ring gear and the sun. This is what allows for the greater gear
ratios, because great dimension differences can characterize these two components.
Brakes
Brakes are mechanical devices designed to slow or stop a machine. Brakes also are intended to
prevent a device from moving after it has been stopped. In wind turbines there typically are two
distinctive brake classificationsaerodynamic brakes and mechanical brakes.
Aerodynamic Brakes
Aerodynamic brakes generally are used in the event of overspeed. Braking during overspeed is
achieved by increasing drag at the tip of the blade, which brings the rotor to a safe rotational
speed. There are many types of aerodynamic brakes, including variable-pitch blades, deployable
blade tips, spoilers, flaps, and brake parachutes. The rotor nevertheless has an effective
aerodynamic area. Aerodynamic brakes therefore generally do not fully stop the rotor; however,
they still drastically reduce the load of the mechanical brake. The standard braking maneuver
typically is to apply the aerodynamic brake and then the mechanical brake; this reduces the
inherent loading on the braking operation. The main problem presented by aerodynamic brakes is
non-simultaneous deployment of the brakes on each blade, which causes aerodynamic
imbalances that can be catastrophic [3].
13
report do not include this parameter, and only represent emergency braking maneuvers that are
performed solely by the mechanical brake.
Mechanical Brakes
The most common mechanical brake used in wind turbines is the disc brake. Although it also can
be seen on the low-speed shaft, in most designs it is located on the high-speed shaft because the
torque loads are much less. The brake is composed of a steel disc that is rigidly fixed to the
braked shaft. A set of calipers are fixed to the frame of the shaft is to be stopped. The calipers
apply enough force to the disc to stop the shaft. The calipers on wind turbines generally are
failsafe, thus the stopping load is applied by springs and the calipers are opened using hydraulics,
and if there is a failure in the hydraulics then the brakes are applied [5].
The brake is a frictional device, therefore it can induce a nonlinear torque increment as it is
applied to stop the turbine. There also is the possibility of inducing additional vibrations to the
gearbox and drivetrain. The modeling of the brake generally neglects these two phenomena, and
is represented by a linear torque increment against the rotational motion. This increment peaks at
a defined maximum braking torque. The rate at which the peak is reached is of great relevance
because it directly dictates the magnitude of the torsional forces induced on the drivetrain to
oppose its inertial behavior.
Generator
Generators are devices that transform mechanical energy into electrical energy. The electrical
power is produced by passing a conductor through a uniform magnetic field at a right angle to
the lines of electric flux. The voltage generated is a function of the velocity, the conductor
length, and the magnetic flux density. The magnetic field used by the generators is obtained by
the use of electromagnets or permanent magnets [3].
The induction generator is the most common generator used in the wind industry. This largely is
because it has a simple configuration and a low price. Its main disadvantage is that it does not
use permanent magnets, thus it has to be connected to the grid to be capable of producing
power [7].
Control Systems
The purpose of control systems is to keep the wind turbine operation within permissible ranges
especially in the event of high winds. The parameters to be controlled are rotational speed,
torque, and rotor thrust. Basic aerodynamics concepts such as leading edge, trailing edge, lifts
and lift drag curve, and flow separation are helpful in understanding the information provided in
the following section. Some of these concepts are explained in the Appendix (Aerodynamic
Simulation, below).
14
Pitch Control
Pitch control is based on the ability to vary the angle of the blades, which consequently changes
the angle of attack of the blade. This allows the rotor to have the optimum angle of attack for
different wind velocities, thereby putting the turbine at its optimum performance for a greater
range of winds. In the case of high-wind control, an implementation of pitch control is pitching
towards feather; meaning that the leading edge of the blade is oriented towards the wind. As a
result the angle of attack is reduced, and the lift or driving force generated by the airfoil also is
reduced. The main disadvantage of pitching towards feather is that a large angle of rotation is
required. This results in a higher-cost system and a slower response toward fast changes in wind
conditions. The major advantage of pitching toward feather is that it has the lowest rotor thrust in
high winds.
A different application of the pitch control is pitching towards stall. In this case, the blade is
rotated in the direction opposite that of pitching toward feather. This increases the angle of attack
away from its optimum, as can be seen in the lift and drag force versus angle of attack graph
supplied in the Appendix. With a great angle of attack the drag is increased and the lift force is
reduced, thus controlling the angular velocity and the amount of torque produced [3].
Stall Control
In stall control, the angle of the blades is fixed to the optimal angle of attack for the rated wind
speed. The angular velocity of the rotor is kept constant by using the torque load of the
generator. When the wind speed increases, the inflow relative to the blades fixed position
changes, and the angle of attack increases. Ultimately, this can cause the flow to separate from
the blade, increasing drag and reducing lift. This reduction of lift and increase in drag reduces the
power output of the rotor, thereby inherently controlling power and torque. The advantage of this
design is that it is a passive power-control method and therefore does not involve any moving
parts. This makes it a more affordable and popular design in the industry. This system also can
react much more quickly to sudden changes in wind speed. The main disadvantage is that the
system has greater levels of thrust because the blades cannot be pitched towards feather. The
other components of the turbine therefore must be more robust to sustain the greater loads [3].
The main advantage of the stall-controlled turbine is its simplicity. Due to the small number of
control parameters, the influence of the control system can be ruled out as the cause of the
observed failures, and a comparison between a great number of subjects can be established. The
ability of the stall-controlled turbine to react almost immediately to changes in wind velocity
reduces the torsional load spikes that are observed in a pitch-controlled system. The main
disadvantage of the stall-control system is the greater thrust loads from the rotor that are imposed
on the drivetrain. To reduce the transmission of thrust loads to the other drivetrain components,
great care is taken in both the design and during assembly. This is achieved by using a low-speed
15
shaft bearing designed to accept the thrust load, assembling it on the shaft with a thrust preload,
and designing the bearing location on the low-speed shaft as far forward (upwind) as is practical.
Gears
Gears are mechanical components that have the ability to transmit motion from one shaft to
another. The transmission of motion might not be uniform, and it can include changes in direction
and torque. More specifically, gears can be described as toothed wheels which typically are round
(but not necessarily) [8]. Gears have evolved from rolling cylinders, to wooden wheels with pegs,
to the common gears used today which have specially shaped teeth that are shaped or cut into
metal wheels. Figure 10 shows wooden-pegged wheels that were used as gears in earlier eras.
out r
mv = = in (2)
i n rout
The positive and negative sign accounts for the internal and external gear sets; an external
set reverses the direction of rotation between cylinders, and an internal set maintains the same
direction. The torque ratio or mechanical advantage ma is the reciprocal of the velocity ratio
(equation 3).
1 r
ma = = in = out (3)
mv out rin
Thus, the gear set essentially exchanges velocity for torque or vice versa. For calculation
purposes, the magnitude of the velocity ratio is expressed as the gear ratio [6].
16
mG = mv or mG = ma for mG 1 (4)
This exchange of torque and angular velocity is very relevant to the loading condition inherent to
the drivetrain. Due to the high gear ratios of the gearboxes implemented in the wind turbines,
small changes in the input velocity produce great changes in the velocity of the generator.
Similarly, small changes of the applied torque at the output of the gearbox result in greater
changes in torque for the initial stages of the gearbox.
Gear Types
Over the years, many types of gears have been developed based on the industrys needs. These
mainly can be classified into parallel axis and non-parallel axis gears. The parallel-axis type can
be classified into internal and external spur and helical gears. The non-parallel axis can be
classified as bevel, hypoid, spiroid, and helicon, among others. The main gearing used in the
wind-energy industry are spur and helical, therefore these are the only ones examined here [8].
Spur Gears
Spur gears formerly were the type most commonly used in industry due to their simple
manufacturing and reduced costa consequence of the alignment of the teeth with the axis of
rotation. The profile of the tooth is designed to maintain constant ratio from one gear to the other
gear. This keeps the tangential velocity from the pinion equal to the gear, ensuring a smooth and
quiet operation. There are several types of tooth profiles that satisfy this condition, although for
industrial applications the most common is the involute profile.
The involute profile follows the shape of an involute curve, and can be visualized as the curve
described by the end point of a taught string uncoiled from a cylinder [6]. This profile is used
widely in the industry because of its simplicity, which eases its manufacture. Additionally, it is
not as sensitive to the center distances between gears, which reduces manufacturing tolerances of
the overall gearing system. The involute profile can operate under greater loads than those of
other profiles and, together with the tolerance to center distances noted above, is capable of
absorbing the deflections of other components of the system. [9]. Figure 11 shows the shape of
the involute curve, originating from the base circle and progressing (always) perpendicular to the
tangent of the taught uncoiling string.[6].
17
The involute profile is the tooth profile typically used in the wind-turbine industry. The
SIMPACK software used to build the progressive models has built-in gear modules capable of
simulating tooth-contact interaction. The integration of the tooth behavior into the multibody
system models is the main difference between the torsional models and the more comprehensive
models.
Helical Gears
Helical gears have teeth that are cut at an angle relative to the axis of rotation. The advantage of
this modification is that these gears have more teeth in instantaneous contact; they also share the
load, operate more quietly, and can withstand greater loads. The main disadvantage of this
modification is that the gears produce an axial load due to the normal loading of the tooth. This
can be compensated for by either using bearings that support axial loading or using a herring
bone configuration, in which the gear contains two helixes with opposite angles that cancel the
axial forces [8].
Helical gears very commonly are used in wind turbine gearboxes. Planetary stages and parallel
stages, for example, are composed of helical gears. The axial forces generated by the helical gears
could be responsible for the premature failure of bearings throughout the gearbox, among others
things. This is regardless of whether bearings capable of withstanding axial loading are used.
18
Pitch circletypically divides the tooth profile in two sections, generally at five-ninths of
the height of the tooth (in accordance with the American Gear Manufacturing Association
specifications for standard tooth proportions). The portion of the tooth extending from the
pitch circle is called the addendum and the section below is called the dedendum.
Line of actiona line described by the points of contact of the gear tooth as it rotates.
The line of action is tangent to both base circles of the gear set and passes through the
pitch point.
Pressure anglethe angle between the tangent of the pitch circle and the line of action.
Polishing is a mild type of wear whereby the surface roughness asperities of the contacting
surfaces are reduced and the surfaces become smooth and mirror-like [10]. The process rarely
results in failures (although, if allowed to progress, it can result in failure). Polishing generally
occurs in low-speed applications in which the elastohydrodynamic lubrication is insufficient.
19
Abrasion
Abrasion (see Figure 15) is caused by particles that have hardness near to or greater than the
hardness of the gear and which are suspended in the oil film. To create the abrasion the particles
also must be larger in diameter than the oil film. Abrasion appears as small grooves that are
carved outwardly from the axis of the gear. It can be prevented by using a filtering system that
eliminates larger particles from the system [8].
20
Surface Fatigue
Surface fatigue is generated by the loading and unloading of the tooth face. The failure originates
under the surface of the tooth when the endurance limit of the material is exceeded. Surface
fatigue generally is not catastrophic during its early stages, although it typically progresses and
ends in catastrophic failure, depending on the loading conditions [10].
Micropitting
Micropitting (see Figure 17), also known as initial pitting, is characterized by the formation of
small pits on the surface of the tooth. The size of the pits ranges from 0.38 mm to 0.76 mm
(0.015 in to 0.030 in). The failure originates at localized overstressed areas and subsequently is
followed by a redistribution of the load. This redistribution can result in the cessation of the pit
production. The subsequent operation of the gear generally results in the polishing of the pitted
area, improving its appearance. In many cases, if the tooth surface geometry is disrupted by the
initial micropitting, then the pitting progresses and evolves into macropitting and tooth breakage.
21
Macropitting
Macropitting (see Figure 18), also known as destructive pitting, is a progression of micropitting
and results in the destruction of the tooth profile. It usually is a consequence of poor load
distribution and high Hertzian stress [8]. Its appearance is similar to that of micropitting, but it is
larger and has a more irregular shape [10].
Spalling
Spalling (see Figure 19) is similar to macropitting except that the damage is quite extensive,
more irregular, and quite shallow. It is more prominent in materials of medium hardness,
although it also can occur in highly loaded fully hardened materials [10].
22
Crushing
Crushing failure mode occurs in surface-hardened or case-hardened gears, when the endurance
limit is exceeded below the hardened region. The failure usually is in the form of a crack that,
once it originates (typically in the interior where the metal is softer), tends to travel towards and
through the surface-hardened portions of the gear. [10]
Plastic Flow
Plastic flow (see Figure 20) occurs from the yielding of the surface and subsurface of the tooth
caused by high-contact stresses. Generally, gears manufactured using softer material are prone to
this type of failure; however hardened gears under heavy loads also succumb. [10]
Fracture
Fracturing (see Figure 21) can be produced by fatigue, tooth bending, or overload. It usually
results in catastrophic failure and its progress is more rapid than the mechanisms discussed above
[10]. This type of failure results in the immediate loss of serviceability, and usually occurs with
little or no warning.
23
Bending fractures generally originate at the root of the tooth; cracks start and propagate,
breaking the entire tooth or a portion of it. This type of fracture can show signs of fatigue such as
beach marks or fretting corrosion and a point having a jagged appearance, which is the last point
to break.
Overload breakage typically has a fibrous or stringy appearance that generally shows evidence of
being pulled apart abruptly and rapidly. No signs of fatigue are present [10].
Bearings
Bearings are devices that allow constrained motion, generally rotational motion. Over time, these
devices have evolved to a variety of configurations that are dependent upon their application.
The most common bearing used in the wind industry is the rolling element bearing. These
bearings are chosen due to their low friction and high load capacity. The wind industry, however,
now is moving toward the use of roller bearings. This is because, for larger bearings, roller
bearings have a higher capacity and are more cost effective. In the case of axial loading, the use
of tapered roller bearings usually is implemented.
Surface fatigue failure (see Figure 22) generally is progressive and can be classified into
micropitting, macropitting, and spalling. Typically the raceway fails first with respect to the
other components of the bearing under surface fatigue. The bearings give an audible indication
of pitting. Pitting increases the vibration of the system, and can progress and result in spalling or
fracture of the rolling elements. Spalling produces a large amount of debris [6].
24
Lubrication failure is the result of the absence of the required lubricant film thickness that is
needed to prevent the contact of the rolling elements and the raceways. The lack of the required
film thickness results in metal-to-metal contact and causes the bearing to overheat. The symptom
typically is discoloration of the roller elements, rings, and cages. In many cases, high
temperatures also can degrade or destroy the lubricant [33]. Overheating (see Figure 23) results
in the loss of bearing-material hardness, which eventually causes failure. In the case of wind
turbines, many of the bearings operate at low rpm, which consequently induces wear due to a
loss of film thickness or complete loss of elastohydrodynamic suspension. This mechanism,
however, is not as catastrophic and does not progress as rapidly as overheating.
25
produced by this failure mechanism inherently causes the abrasion of other components of the
gearbox.
There are characteristic loading conditions that drive this surface-fatigue failure mechanism in
the bearing, including misalignment and reverse axial loading. Other parameters that influence
the failure of bearings can include the bearing fitting either loosely or tightly with the housing,
which can result in fretting.
Nevertheless, gear failures also occur independent of bearing failures, although not as
commonly. The most common gear failures in the industry include wear from poor lubrication
(generally seen in the planets due to their low rpm), abrasion, and surface fatigue initiated by the
debris generated from bearing failures. Other common causes of gear failures include tooth-
interference, poor tooth load distribution produced by misalignment, and changes in center
distances due to gearbox housing compliances.
The models developed in this report cannot accurately predict any failure related to lubrication
such as wear; however they can predict the loading conditions produced from misalignment,
changes in center distances, backlash, and high loads generated from the interaction of
dynamical components. The prediction of these parameters reveals important information that
can be used for the life-analysis of the different components. Table 2 shows the capabilities of
the models with respect both to load-case prediction and to the models progressive complexity.
The model of highest complexity presented in this table is a finite element model. The advantage
of this model is that it can generate loading conditions, as well as analyze internally the stress
and life-reducing parameters of the gears and bearings. The main disadvantage of this approach
is the greater computational time due to the great number of degrees of freedom.
26
FAST_AD is operated by means of a number of input files that describe the properties of the
turbine as well as its aerodynamic characteristics. The main file or primary file describing the
turbine has a series of flags that can be changed to true or false to match the simulated turbine.
Parameters of note described in the primary file include simulation controls, which control the
numerical parameters of the simulation; the turbine control, which describes the controlling
parameters of the turbine such as start and stop times for the generator and the brake as well as
for the control of pitch angles; turbine configuration, which describes masses and physical
dimensions of the turbine stiffness and gear ratios; and finally parameters which describe two
generator models. Additionally, this file contains the required names for other input files that the
code must use as the simulation progresses.
FAST can incorporate user routines for the controls of the turbine and to describe the generator,
in case the predisposed models provided by FAST do not replicate the behavior of the turbine
closely enough. The code also handles routines for simulating turbulent wind input to the rotor
and handles routines to determine aerodynamic loads.
The motion of the body is described by equations which refer to an inertial frame. An inertial
frame is a coordinate system that translates with uniform velocity and constant orientation with
respect to the stars [13]. The equations describe the motion of a body with respect to time, and are
constructed in relation to the number of degrees of freedom of each body and the overall system.
Two types of coordinates can be used to create the equations of motionabsolute and relative.
Absolute coordinates relate each individual body to the inertial frame, which results in
consistently having the maximum number of dimensions in the equations of motion.
Consequently, the time integration is heavier and more time consuming. Relative coordinates
rely on a kinematic tree structure in which the degrees of freedom of each body are specified
with respect to the previous body, but only the first body is related to the inertial frame. This
27
approach results in equations of motion with minimal coordinates, thus making the time
integration more efficient.
Figure 24 shows both types of coordinate systemsthe absolute coordinate system is on the left
side, and the relative coordinate system is on the right side. [14] In short, the multibody-system
principle decomposes the system into free-body diagrams, enabling the degrees of freedom can be
depicted. The kinematic loops are defined describing motions in relation to adjacent bodies by way
of joints, constraint, and forces. The equations of motion are formulated and integrated over time.
This software is operated via a graphical user interface (GUI). The user inputs parameters
describing the system to be simulated, such as mass and inertias of each body, and then can create
kinematic loops by applying the pertinent joints, constraints, and forces. SIMPACK creates the
equations of motion internally and allows the user to choose different options for the time
integration. To obtain visual animations of the interacting system, the user also can input three-
dimensional primitives which describe the geometries of the interacting bodies.
SIMPACK has a force element library containing specific force elements tailored for different
areas of the industry, including specific elements designed for the drivetrain simulation
(explained in more detail in the following section).
The SIMPACK software has parameterization capabilities enabling storage of the input
parameters of an arbitrary system on a database that contains a number of input files. These input
files can be modified and stored separately, depicting a particular condition for a particular case.
This feature simplifies the iterative process of data comparison and allows the system to be re-
dimensioned to other similar systems. [14]
28
FE:12
Body 1 Body 2
Revolute Joint
Input parameters of the force element include the joint identification, from which joint states are
used for the calculations. The springs torsional stiffness and damping, the pre-stress moment,
clearance, and angle of zero torque also can be defined. In the case of nonlinearity, a function
describing the nonlinear behavior can be used for the stiffness and damping parameters. The
output of the force element can be seen as the joint forces on which the element is acting.
The input parameters for the force element are: the orientation of the axis where the input and
output moment is applied, the label of the connecting joints, the gear ratio, and the stiffness and
damping of the system, as well as a pre-stress moment. The outputs of the force element are the
torques from both shafts and the difference in angular rotation.
29
flank, as well as the face-width of the pairing gears. The forces are calculated from the
penetration of the contact points into the tooth profiles, which lay at the line of action. The force
element also is capable of reproducing the interaction of gears of different materials and backlash
for reversal of direction.
The force element uses standard gear-wheel physical parameters described from a three-
dimensional gear-wheel primitive. The primitives are limited to involute tooth geometry;
although they can account for spur or helical gears as well as internal and external gears. Some
parameters describing the primitive are normal module, profile shift factor, addendum,
dedendum, backlash factor, number of teeth, face width, and bore diameter. The force element
then requires material properties of the gear wheel, including stiffness model, friction model, tip
relief factor, and damping coefficient. The main limitation of the force element is that the axes of
the pairing gears must be parallel.
The output of the force element includes: rotational velocity, difference in angle, difference in
velocity, contact region, contact stiffness, total circumferential force, axial force stiffness,
damping force, friction force, pitch points, angle of attack, dynamic transmission error,
penetration velocity, and damping.
The shaft is subjected to pure torsional loading, and no axial bending or direct shear load are
present. The cross section of the shaft remains a plane and perpendicular to its axis. The material
of the shaft is homogeneous, is isotropic, and obeys Hookes law. The stresses realign on the
elastic limit and its behavior is purely linear. The interaction of the boundary conditions of the
bar doesnt influence the behavior of the bar. Initially the shaft is straight. [6]
30
Hookes Law
The mechanical behavior of any material can be described by a relation between stress and
strain. The described curve varies its shape depending on the material properties such as ductility
and hardness. For all materials, the initial stages of this curve have a quasilinear behavior. This
quasilinear behavior can be said to be linear for all materials; this generalization is known as
Hookes law. The magnitude of this quasilinear stage is dependent also on the material
properties. Hookes law can be represented by the following equation.
= E (5)
Where is stress, E is the proportionality constant called the elastic modulus or Youngs
modulus. is strain and dimensionless, therefore E has the units of stress. Graphically E can be
said to be the slope described by a straight line from the origin to a point A on the stress versus
strain curve. Point A lies at the end of the quasilinear portion of the curve (see Figure 27).
Brittle material
A Ductile material
A
Some organic materials
A
Figure 27. Stress versus strain representative curves [12]
The linearization noted above is of great relevance here, because all the parameters used to create
the progressive models are described with a linear response.
31
Y
xy
Hookes law can be extended to this type of loading, creating a liner relationship between shear
stress and angular deflection. This can be represented by the following equation.
= G (6)
Where is the shear stress, is the angular deformation, and G is the proportionality constant
(shear modulus of elasticity). [12]
Gr1
= (7)
l
Where is the shear stress, is the torsional deformation of the bar, G is the proportionality
constant (shear modulus of elasticity), r1 is the radius of the cylindrical side of the element, and l
is the length of the bar. Because , G, and l, are constants, the value of the shear stress () varies
directly with the radius. The applied torque can be calculated by integrating over the cross-
sectional area, yielding the following.
r
T = r1 dA (8)
0
The shear stress changes linearly from the axis of the bar, so the following is a constant.
r1
32
r r
T =
r r
2 2
r1 dA = r 1 dA = Ip (9)
0 r1 1
0
1 r1
The integral
r 2
r
0
1 dA
is substituted by the variable Ip, which is called polar moment of inertia and is the ability of an
object to resist twist; in this case it is for a circular solid rod.
d 4 r 4
Ip = = (10)
32 2
The units of the angular displacement are in radians. [9] The general equation for the calculation
of the stiffness is as follows.
T
k= (12)
If the previously obtained deflection is applied, it shows that the stiffness is solely dependent on
the geometry and material description. It can be represented by the following equation.
TI p G I pG
k= = (13)
Tl l
kt
n = rad/sec (14)
Im
33
Where n is the lowest free vibration frequency in radians per second, kt is the torsional stiffness
in N-m/rad, and Im mass moment of inertia is in Kg - m2. If the natural frequency is provided in
Hertz, it can be changed to radians per second using the following expression.
n = 2f n (15)
mx + cx + kx = 0 (16)
Where c is the damping coefficient and has the dimension of force per unit of velocity [16]. This
parameter is the most difficult to estimate; the other parameterssuch as inertia and stiffness
can be calculated easily or measured statically. [17]
A good way to describe the damping behavior of a system is by using the damping ratio . This
is the ratio of the damping coefficient with respect to the coefficient of the same system critically
damped. A critically damped system is a system in which no oscillation can occur independent of
the magnitude of excitation. The damping coefficient for a critically damped system can be
calculated using the following expression.
c = 2 Imk (17)
Where I is the inertia of the body and k is the stiffness of the connecting spring. Thus, a critically
damped system has a damping ratio of 1.
Logarithmic Descent
When dynamic testing is possible, a good way to approximate the damping ratio is by using a
logarithmic descent (Figure 29). This is defined as the natural logarithm of the ratio of two
successive amplitudes. It can be obtained from the following expression [17].
x(t )
= ln (18)
x(t + P)
Where P is the period and x is the amplitude of the oscillation. Substituting the characteristic
roots for an under-damped system into the previous equation and performing some manipulation
results in the following expression.
2
= (19)
1 2
34
= (20)
4 2 2
A plot of x versus t generally contains some measurement error. For this reason the previously
described method is modified to measure two peaks n cycles apart. The sequential peak values
can be denoted by B1, B2, etc.
Then
B B B B B
ln 1 2 3 n = ln 1
B2 B3 B4 Bn+1 Bn+1
or
B1 B B B B
ln + ln 2 + ln 3 ln n = ln 1
B2 B3 B4 Bn+1 Bn+1
thus
Bn
+ + + = n = ln
Bn+1
or
1 Bn
= ln (21)
n Bn+1
35
Equation 21 can be used to find the damping ratio, and the damping coefficient can be calculated
using the following expression.
c = 2 Ik (22) [17]
Pl 3
y max = (23)
3EI a
Where P is the applied load at the tip of the beam, l is the length of the beam (in this case the
height of the tooth), E is the modulus of elasticity, and Ia is the area moment of inertia (in this
case it is assumed to be a rectangular cross section). The area moment of inertia for a rectangular
cross section can be determined using the following expression.
bh 3
Ia = (24)
12
Where b is the width of the cross-sectional area of the beam (in this case the operational face
width of the gear tooth), and h is the height of the cross-sectional area (in this case the tooth
thickness).
Generally, the gear mesh stiffness is calculated from the angular deflection of the gear per unit of
torque. This means that the calculated tooth deflection has to be changed to angular deflection.
Because the linear beam theory assumes that there is no change in length of the beam, and given
that the angle of deformation is small, the overall angular deflection of the gear (see Figure 30)
can be calculated using simple trigonometry.
36
The width of the beam is taken to be the effective face width or the smallest face width of two
meshing gears in a gear set. The thickness of the idealized beam can be related to the tooth
thickness which also can be obtained from the AGMA full-depth gear tooth specification.
Blade Inertia
For the blades moment of inertia the additive property is employed. This states that for a rigid
body consisting of N point masses mi with a distance ri to the axis of rotation, the total moment
of inertia equals the sum of the inertias of the point masses. It can be expressed by the equation
below.
N
I m = mi ri 2 (25)
i=1
To perform this equation the blade must be converted into point masses. This is done by
subdividing the blade along it length, calculating the mass for each section, and assuming that
the mass is effective at the center of the subdivided section. Figure 31 shows the discretization of
a blade into mass-point form.
37
This approach assumes that the blade behaves like a completely rigid body, which might not be a
true assumption for certain load cases.
Depending on the gear ratio, a new relation for the stiffness can be derived from the principle of
potential energy. To rescale the inertias to a single lumped inertia, the principle of kinetic energy
can be applied.
Figure 32 (below) is a graphic representation of a simple gear system. It is assumed that the
shafts have no inertial properties and so only their stiffness is considered. The gear ratio of the
system is given by the following.
n
1
Hence it is n . [5]
38
In Figure 32, representative two-stage drivetrain inertias are represented by J and stiffness of the
shaft is represented by k. [5]
When the system is compounded the shafts are represented by a single shaft, and a new effective
stiffness is calculated by scaling the original values of stiffness with the square of the gear ratio.
In the same manner, the inertial behavior of the system also is adjusted by scaling it with the
square of the gear ratio [5]. Figure 33 is a graphic representation for the equivalent model of the
drivetrain.
Figure 33. Equivalent one-stage drivetrain with single stiffness and inertia [5]
The reduced stiffness can be calculated by combining stiffness with the following expression.
n 2 k1k 2
k eq =
k1 + n 2 k 2
39
Turbine Description
To create the baseline model, a physical turbine with a number of particular characteristics was
chosen. The turbine selected is a turbine representative of the current industry standard. This
turbine also served as a basis for the creation of the simplified models, and their validation by
means of data acquisition.
The turbine is a horizontal axis, three-bladed machine with rated power in the range of 750 kW.
This is the smallest turbine that can maintain the overall configuration characteristics of larger
turbines (1.5 MW and 5 MW). The turbine has a rotor diameter size of 48 m and a tower height
of 55 m.
The drivetrain follows the prominent modular configuration previously mentioned, and is
composed of a low-speed shaft, gearbox, brake, high-speed shaft, and generator. The gearbox
follows the typical configuration of a single planetary stage and two consecutive parallel-axis
stages with helical gears. The location of the brake on the high-speed shaft allows the isolation of
the behavior of the low-speed shaft and gearbox with respect to the generator and the rest of the
turbine. The generator used is a double induction generator having the capability of switching to
a different number of poles, thus allowing its operation at two separate constant speeds. The
simulation of this transition is outside the scope of this report.
The turbine is an upwind stall-controlled type. This configuration eliminates a number of control
variables that could greatly influence the loading condition of the drivetrain, due to the high
sensitivity to the parameters involved in pitch-regulated control systems.
The physical presence of the turbine and the cooperation of wind-turbine site owners also are
valuable components, because they simplify the data-acquisition process.
40
of the gear ratio of the drivetrain is included by calculating an effective inertia that is proportional
to the square of the gear ratio (see Effective Inertia and Stiffness, above).
The bodies are connected to the reference frame with only one degree of freedom (rotation),
giving the overall system only one DOF per body. This connection has unknown properties such
as stiffness and damping coefficients, and is assumed to be without mass. These parameters are
determined by the use of experimental data.
Although this is the simplest model that was built, it is of greatest importance because it serves
as the basis for the input and validation of the subsequent stages. Additionally, this simplified
drivetrain model is used by many aeroelastic codes which are used to generate the loading
conditions of the drivetrain. This configuration serves as further validation for the aeroelastic
models that already have been validated.
The system with the topology shown in Figure 35 (see Figure 36) was implemented for the semi-
static validation. Note that the body representing the generator is constrained to the reference
frame with zero degrees of freedom. This simulates the rigid application of the brake on the
high-speed shaft. In this configuration the interaction of the generator can be ignored and omitted
from the equations of motion.
41
Generator
0 DOF
Rotor
FE:12
Y
X
Figure 36. Topology of constrained model
As shown in Figure 37, the constraints imposed on the generator are removed and only rotation
around the X-axis is allowed. This system has two degrees of freedom, therefore the behavior of
both main components must be monitored to obtain the correct dynamical behavior of the
connecting spring. Generator
Rotor
Y FE:12
X
Figure 37. Topology of unconstrained model
The first step of the validation involved the acquisition of the physical parameters of the system.
For this simple model, the only requirements were the mass and moment of inertia of the rotor.
42
The bulk masses for the blade and hub were acquired, [27] and a typical mass distribution for a
750-kW blade was scaled linearly to meet the bulk mass of the blade. The mass distribution also
was scaled to match the actual length of the blade. Using the specific characteristics of the mass
properties, the blade was discretized into point masses along its length. The mass moment of
inertia for each point mass was calculated and summed to obtain the overall inertia of the blade.
This approach assumes that the rotor is entirely rigid, and neglects the effects of the flexibility of
the blades. The model has only one degree of freedom, therefore the only inertia of interest is the
rotation around the axis of the rotor (see Blade Inertia (above) for more detail).
To obtain the stiffness and damping, acceleration measurements were recorded on the turbine
itself. These measurements reveal the overall frequency of the drivetrain as well as the rate of
damping of the system. The data collection consisted of performing mechanical braking
maneuvers to induce excitation to the drivetrain. The normal braking procedurethe deployment
of blade tipswas overridden, and the mechanical brake was applied manually (see the
drivetrain section (above) for a description of normal braking maneuver). When the mechanical
brake is applied, the entire drivetrain undergoes torsional loads to oppose the inertial forces of
the rotor. Consequently, the drivetrain coils and uncoils like a spring. The desired information
from the system can be interpolated from this response.
Acceleration was measured on the gearbox housing because it is joined to the bedplate by
elastomers and a stronger excitation signal was captured than if it had been collected on a
component rigidly attached to the main frame. The data analyzed was collected on the top of the
gearbox housing because it provided a horizontal flat surface where accelerometers could be
easily mounted. To enhance the signal collected, the accelerometers were located at the point
farthest from the axis of rotation of the input shaft. The acceleration data was collected by
uniaxial accelerometers placed in the vertical direction to prevent tower sway movement from
being recorded in the data sets. The data collected was scaled so that the acceleration information
collected away from the axis of rotation subsequently could be translated into angular
acceleration, and could be compared directly to the results of the dynamical simulation.
No existing experimental data having the same parameters as the data collected was found.
Therefore no comparison with any existing validated data could be made. The response expected
from the collected data should have resembled the behavior observed in the final stages of
specified braking maneuvers, which can be found in AGMA standard 6006. This existing
braking maneuver data, however, depicts torque behavior with respect to time and does not use
acceleration as a parameter.
43
The data collected nevertheless reveals the overall frequency of the system, because it is a
parameter that is easy to capture. In the case of the damping the amplitude of the excitation
presents a secondary frequency, which can be seen in the figure above as a red line and which
raises doubt with respect to the validity of the data. Regardless of the precision of the data, this
report depicts a method to validate the progressive models generated. Moreover, the values
obtained from the validation of the models correlated quite well with the theoretical calculated
values. This leads to the conclusion that the approach followed and the data obtained were correct.
The acceleration recorded is characterized by strong point excitations that diverge in great
magnitude with respect to the mean acceleration value. These slender excitations contain a single
data point for each peak regardless of the high frequency (1,000 Hz) of data collection.
Consequently, a minute discrepancy at the time of the excitation and the data collection creates a
change in the recorded response. This can be seen as a much lower frequency influencing the
magnitude of the excitation. This phenomenon is produced by a small offset in the frequency of
the system and the data-collection frequency. In Figure 38 this frequency is shown as a red line.
To elaborate on this phenomenon, a pair of graphs is presented in Figure 39. The graph on the
left depicts an arbitrary sinusoidal wave. The vertical lines represent the point in time in which
the data is collected. The red dots mark the point where the sinusoidal response crosses the data
collection; therefore, the dots depict the magnitude of the data collected for that point in time. If
the data points collected were joined, then the response obtained would be that described by the
red line in the graph on the right side of Figure 39. It shows that a secondary frequency bounds
the amplitude of the recorded response. This secondary frequency is depicted by the green line in
the right-hand graph. This essentially is the same phenomenon as seen in the experimental data
collected, although in that case it is produced by the short period of excitation.
44
The effect of nonlinear behavior can be seen in the experimental data, because the initial
frequency recorded is not the same as that at the end of the data collection. This is due to the
stronger influence of clearances with respect to smaller amplitudes of excitation. The slight
changes in frequency consequently alter the secondary frequency observed.
The gaps between the excitations reflect the time in which the internal clearances of the gearbox
such as backlash and bearing are absorbed. This occurs for every cycle, because the loading is
fully reversed once the system has been excited.
From the data presented above, the frequency of the periodic peak excitations was recorded as
1.1 Hz. This value was used in combination with the previously mentioned rotor inertia to
calculate a simplified rotational vibrating spring mass system. Using the assumption that the
system behaves as a free vibrating system, a theoretical stiffness of 6.25E+07 Rad/Nm was
calculated and input into the Stage 1 model shown in Figure 34 (see Torsional Free Vibration;
Torsional Free Damped Vibration). The simulation was performed and the frequency obtained
by the simulation had good concurrence with the experimental response (see Figure 40).
The reduction in the magnitude of the excitation is a consequence of the damping of the system.
Due to the particular characteristics of the data recorded a trend curve could not be fit. This is
because the points of interest are the greatest in magnitude of the secondary frequency and there
are not enough points to result in a converging curve.
The reproduction of the secondary frequency was achieved by changing the time integration
characteristics of the simulation; more specifically, by increasing the size of the integration time
step. This produced a response similar to the point excitation and the secondary frequency
observed in the field, although the shape of the simulated curve resembles a smooth sinusoidal
wave.
A theoretical damping coefficient was calculated by assuming that the overall system behaved as
a free damped vibrating system. The result is a simulated response that closely follows the field
data recorded.
45
When the damping and the stiffness were tuned to match the recorded data, the simulation time
step was increased to obtain the representative form of the oscillating system. The values
obtained for the stiffness and damping are used as guidelines for the validation of the rest the
progressive models. Figure 41 depicts the simulated response with the refined time step. This
case uses a time step of 50 steps/sec.
Figure 41. Stage 1 simulated response with refined integration time step
46
This approach accounts for the torsional compliances resulting from the bending and contact
deflection of the gear teeth, as well as torsional deflections of the shafts. The model ignores the
added torsional compliance from the bending of shafts and from bearing deflection [10].
Spring dampers joining both gears were used to simulate gear interaction. These joints lie over
the line of action and are subjected to the tangential forces of the gears [19]. The overall torsional
response of the system was obtained, as well as the response from the internal components in a
dynamical and coupled manner. The SIMPACK force element FE:14 was used as the tangential
spring interaction [21].
The gear bodies were represented by rigid primitives used in the 3-dimensional visualization, and
do not have any influence in the calculations outcome. The shafts were simulated by torsional
spring dampers or FE:12 in SIMPACK, giving the insight of the torsional shaft deflection as a
separate parameter. All the respective torsional inertias from each individual component must be
calculated from the masses and geometries as inputs for the model. It is assumed that there is no
change in direction, and that the connecting springs between gears always will be under tension
(no backlash).
Figure 43 shows the topology of the planetary stage of the gearbox, which is the most complex
stage. One of the requirements of the force element SIMPACK FE:14 is that the input and output
shaft-force elements are defined with respect to an arbitrary common body such as the gearbox
housing. This limits the ability to simulate the planets inertial behavior. To remedy this situation
a massless connecting body was created. This body has the dynamical outputs such as torque and
angular position of the planets. To account for the inertial behavior, three planets with respective
masses and torsional inertias were created and attached to the planet carrier, with only one
allowable rotation. Using this configuration, the rotational behavior of the planet on the planet
carrier is taken into consideration. To transmit the rotational behavior of each planet to the
47
massless body, the rotational degrees of freedom were constrained; more specifically, degrees of
freedom in alpha were constrained.
Figure 43 is a graphic representation of the planetary stage. Due to the limitation of the
SIMPACK FE:14, each gear interaction requires the definition of one housing (red box) and two
SIMPACK FE:14 elements. The first housing represents the planet ring interaction with the
respective gear ratio, as well as the added rotation from the trajectory of the planet. The input
shaft is represented by the planet carrier and the output shaft is represented the by the massless
connecting body 1. The mass and inertias of the massless body are assigned by the constraint of
rotation in alpha. The importance of this setup is that the masses and positions of the planets
influence the inertia of the planet carrier.
On the second housing the planet-sun interaction is represented and, naturally, the sun-planet
gear ratio is implemented to define the force element. The torque and angular velocity of the
input shaft of this section are transferred by a zero DOF constraint between both massless bodies.
Planet
FE:14 FE:14
FE:14
FE:14
Massless Massless
connecting 0 DOF connecting
Planet
FE:14
Planet
Y 0 DOF
0 DOF
Figure 44 is a continuation of Figure 43, expanding the topology to the remaining internal stages
of the gearbox. At the far left, a portion of the sun is represented as a continuation of Figure 44;
although, to avoid overcrowding, the SIMPACK force element FE:14 is not represented. This
figure shows connecting shafts for the different internal stages of the gearbox. Each individual
shaft is represented by two independent bodies. The first section of the shaft is connected to the
previous gear by a 0 DOF connection. The two sections of the shaft are connected with a single
48
axis rotating joint. The state quantities of the joint are used by the SIMPACK FE:12 (torsional
spring) to calculate forces and displacements of the second section of the shaft. The two helical
stages are represented in the same form as the planet-sun interaction.
Intermediate High-Speed
Stage Stage
FE:14
FE:12 0 DOF
DOF
0 DOF
Pinion
Gear
Sun
FE:14
FE:14
FE:14
FE:14
Pinion 0 DOF
DOF
Gear
Y FE:12 FE:14
0 DOF 0 DOF
X
As noted, all internal components are simulated in this stage. This calls for the dynamical
properties of each component. The following sections explain how specific parameters required
for this model were acquired.
Mesh Stiffness
The parameters of the spring dampers connecting each gear wheel or element FE:14 were
substituted by the gear mesh stiffness. The gear mesh stiffness was calculated as a single simply
supported beam, and it was assumed that the thickness of the beam remained constant and
unchanged by the tooth profile. The stiffness of the tooth therefore is constant along the tooth.
The length of the beam was taken to be the height of the tooth including the respective clearance.
This approach assumes the most conservative scenario, because it only takes into account one
gear tooth being in contact. It also assumes that the contact occurs at the farthest point of the line
of action, instead of acting at the pitch diameter.
The force applied to the tooth for the stiffness calculation is the tangential force of the gear pair,
which is tangent to the pitch diameter and perpendicular to the tooth. This approach does not
account for multi-tooth contact or for changes in the contacting number of teeth; hence the
stiffness remains constant and linear. The model also neglects to consider the existence of
backlash.
Shaft Stiffness
The shafts were modeled as ideal steel circular beams. The influence of steps and stress risers
was neglected, therefore an average or representative diameter had to be considered to be able to
49
generalize the behavior of particular sections of the shaft. It was assumed that there was no
torsional deflection on the shaft under the section braced by gears. The entire defection of the
shaft occurs on the internal distance between the gear pairs; this distance is called the effective
length. This can result in fairly high stiffness because in some situations the effective length of
the shaft or the section undergoing torsion can be quite short. This assumption ignores the
relatively large clearances that characterize key and spline connections between gears and shafts.
These clearances can result in a nonlinear load deflection characteristic. [10] The shafts were
subdivided into two equal sections, represented by two independent bodies connected by the
spring element having the calculated effective stiffness.
Inertias
Inertias were calculated from the physical properties of the gearbox components. The rotor
inertia was calculated in the same manner as was done in Stage 1. The shafts were assumed to be
solid cylinders and all shafts were subdivided into two equal but separate sections. The inertias
were calculated for each independent section of the shaft. The shafts were joined by a single
DOF joint, and the overall inertia and interaction of the shaft was calculated by the simulation
code. The inertia for the respective gears was calculated by assuming that the gears behaved as a
solid cylinder described by the pitch diameter and the face width of the respective gear wheel.
To validate the damping values, a more quantitative approach than that performed on the Stage 1
model was pursued. This approach assumed that the overall system behaved as a single degree of
freedom, torsional free, damped, vibrating system. Using equation 22 (see Logarithmic Descent,
above) the damping coefficient of each individual shaft can be calculated. This equation is
dependent on the stiffness and mass or inertia properties of the single degree of freedom system,
thus a reduction of the actual degree of freedom of the system was required. This was achieved
by assuming that the shaft for which the damping coefficient was to be calculated was the only
section of the system allowed to move. Therefore all the other components of the system were
assumed to be infinitely stiff.
For the inertia component an effective inertia had to be calculated due to the effects of the gear
ratio (see Effective Inertia and Stiffness, above). Again, the approach of using the conservation
of energy was applied to the system, althoughin this caseinstead of increasing the stiffness
of the shaft, the inertia of the rotor was reduced. This is because the stiffness that is input into the
simulation code must be the representative stiffness of the shaft, regardless of the gear ratio. The
damping calculated therefore must be correlated to this stiffness. The effects of the gear ratio
with respect to stiffness and damping are calculated by the software. This approach was used
iteratively for each subsequent shaft; therefore the damping was calculated depending on the
stiffness of the shaft as well as the position along the gearbox.
50
The quantity missingbut which can satisfy the equationis the damping ratio. This was taken
to be the damping ratio from the experimental data, which also was calculated from a logarithmic
descent.
The same braking event as that used in the model for Stage 1 also was induced on the more
complex Stage 2 model. The damping ratio was calculated from the response obtained and then
compared with the damping ratio of the experimental response. The difference was minimized by
scaling the damping coefficients of the simulation, with a 5% scaling reduction of the calculated
theoretical value. This produced an error percentage of less than 3% for the damping ratios.
Figure 45. Stage 2 model simulated response with fine integration time step
Figure 45 depicts the response from the Stage 2 model. A fine integration step of 50 steps/sec
also was used to obtain the response. Notice that, although the percent error between Stage 2 and
the data collected is very small, visual differences in the damping envelope can be seen.
51
Stage 3 uses a built-in SIMPACK force element FE:225. The only requirement of the force
element is that a body-fixed marker in the center of each gear be defined. This marker is required
to establish the reaction forces between the two interacting gears [22].
To account for the force interaction of all components, the ring gear also must be considered.
The Stage 3 ring gear was fixed to the gearbox housing with zero DOF. The force interactions of
the system included planet to ring as well as planet to sun, and the two forces and their directions
were considered in the analysis. The direction and magnitude of the forcesespecially those of
the planetsare of great importance because the force on the tooth changes direction with each
cycle, which increases the fatigue damage. The SIMPACK force element FE:225 can account for
changes in the gear center distance and backlash [22]. For this model, because the joints of the
system have only one degree of freedom, the center distances do not change; nevertheless,
backlash is included.
In Figure 47 the topology of the planetary section is represented for Stage 3 of the analysis. The
main difference between Stage 2 and Stage 3 is the implementation of the new SIMPACK force
element FE:225. This element eliminates the need for the multiple housing representations; one
common housing is used for the entire gearbox model. Note that the only interaction between the
planets and the ring gear, and the planets with the sun, is the force element FE:225. This
eliminates the need to implement the two massless connection bodies shown in Stage 2. The
planets have only one degree of freedom with respect to the planet carrier, and there are no added
constraints to the system. The simulation with the new force element, as well as with all the gear
components, provides a much more realistic representation of the system and its interactions.
52
0 DOF
FE:225
Planet
FE:225
Gear
Ringgear
Planet carrier
FE:225
Ring
FE:225
Planet
Sun
0 DOF
Connection
to shaft
FE:225
Planet
FE:225
Y
0 DOF X
Figure 48 is a continuation of Figure 47, expanding the topology to the remaining internal stages
of the gearbox. The shafts are represented in the same manner as in Stage 2, with the use of
SIMPACK FE:12 as a torsional spring damper. The main difference is that the shafts now share
a common housing with a joint that only allows rotation in alpha. The two helical stages also can
be seen. These are connected to the nearest shaft by a zero DOF joint and SIMPACK force
element FE:225 is between them.
0 DOF
0 DOF
0 DOF
Connection FE:12
from Sun FE:12
FE:225 FE:225
Y
0 DOF
X FE:12 0 DOF
53
Although Stage 3 is more complex it still is extremely similar to Stage 2, making for a good
comparison between stages. The following sections explain the acquiring of the specific
parameters required for this model.
Mesh Stiffness
In this case, the mesh stiffness is calculated in a more accurate manner internally by the
simulation code. The software calculates mesh stiffness using the parameters describing the
gearwheel primitive, together with additional user-supplied material properties. As noted above,
the advantage of using the SIMPACK force element FE:225 is that the system accounts for
multi-tooth contact and the variation in stiffness resulting from the change in number of teeth.
This model also accounts for backlash (see SIMPACK Force Element FE:225, above).
Shaft Stiffness
As in the previous model, the shafts for Stage 3 were modeled as ideal steel circular beams. The
influence of steps and stress risers was neglected. Therefore, to generalize the behavior of
particular sections of the shaft, an average or representative diameter was considered. For more
information, see the Shaft Stiffness section for Stage 2 (above).
Inertias
Inertias were calculated from the physical properties of the gearbox components in the same
manner as was done in the Stage 2 model. The main difference in the inertial calculation is that,
in the previous model, all shafts were subdivided into two equal, separate sections. In this model,
due to the more detailed configuration, the shaft was not necessarily subdivided into equal
sections. This resulted in the need for recalculation of the inertia for the particular sections.
As was done in the previous model, it was assumed for this model that the overall system
behaved as a single degree of freedom, torsional free, damped, vibrating system, and the same
approach of the logarithmic descent was performed (for more detail, see the description of the
Stage 2 model).
The same braking event that was performed in the Stage 1 model was induced in the Stage 3
model. Using the response obtained, the damping ratio was calculated and compared with the
damping ratio of the experimental response in the same manner as for the Stage 2 model. The
difference between the experiment and the simulator response was reduced to a minimum by
scaling the damping coefficients of the simulation. Using a scaling increment of the calculated
theoretical value of 20% resulted in a less than 1.5% error for the damping ratios.
54
Figure 49. Stage 3 model simulated response with fine integration time step
Figure 49 shows the response of the Stage 3 model. A fine integration step of 50 steps/sec also
was used to obtain the response. Note that although the percent error between this stage and the
experimental response is very small, noticeable visual differences in the damping envelope
appear when it is compared with the previous two stages.
55
The addition of the force element required the adding of new markers that describe the
attachment points of the bearings. This is very relevant because the bearings are represented by
three stiffnesses (see Figure 50), and each shaft needs a minimum of two constraining forces.
This situation required a more precise description of the housing and the exact positions of the
bearings. Additionally, the length of the shafts and the position of the bearing on the shaft were
of great importance because they had to match the housing to prevent the undesired prestress
conditions. This was not required on the other models because they were purely torsional models
with rigid bodies.
Constraints also were added to the connection between the input shaft and the gearbox.
Originally only rotation was constrained, here all six DOF are constrained. This simulates the
rigid coupling between the shaft and the gearbox, and reveals how nontorsional loads affect the
internal components of the system.
The diagram in Figure 51 illustrates the additional degrees of freedom for the planetary stage.
Comparing this figure with Figure 47 shows that the overall structure of the model is the same,
with the addition of new force elements and new translational degrees freedom per joint.
56
0 DOF
X,Y,Z
FE:225
Planet
FE:43 FE:225
Gear
Ringgear
Planet carrier
X,Y,Z FE:225
Ring
X,Y,Z FE:225
Planet
Sun
FE:43 FE:43
0 DOF
Connection
to shaft
X,Y,Z
FE:225
Planet
FE:225
Y
0 DOF
FE:43
X
Figure 51. Planetary stage with added degrees of freedoms and force constraints
57
FAST Model Description
FAST Input Parameters
The detailed description of the input parameters for the simulation code FAST_AD required that
a number of parameters be interpolated and calculated. The following sections briefly describe
the parameter-acquisition process.
Blade Characteristics
A number of parameters describing the blades were required, such as: length, cord and twist
distribution, airfoil characteristics and location along the blade, mass distribution, stiffness
distribution, and mode shapes. Due to the lack of some of this information, a similar blade from a
turbine of the same power rating was used as a reference blade to aid in the scaling of the factors.
The length as well as cord and twist distribution were measured from an existing blade. The
reference blades airfoil characteristics and respective locations also were used. The overall mass
of the blade was obtained from existing documentation. The mass distribution was scaled
linearly to meet the overall mass of the simulated blade. The mass distribution also was scaled so
that the blade length and the center of mass would match the blade. In case of the stiffness
distribution, the same approach of linear scaling was followed.
For the mode shapes calculation, the modal frequency response of the blade was recorded from
the field. This was done by manually inducing first and second mode in the flap direction, and
first mode in the edge direction. The response was recorded using accelerometers. The
frequencies recorded were compared with frequencies from other similar blades for reassurance.
The mode shapes for the blades were calculated from the data described.
Tower Properties
The mass and stiffness parameters were calculated from existing drawings of the towers, which
described the towers overall geometry as well as its wall thickness. The side-to-side frequency
response was measured by exciting the structure with a similar mechanical braking maneuver
(described in the model development section). The required mode shapes for the model were
calculated from this information.
Generator Models
As noted, the turbine to be simulated has a variable induction generator that allows it to operate
at two different constant speeds (see Figure 52). This is achieved by changing the number of
poles from six to four for operation at 1,200 rpm and 1,800 rpm, respectively. As stated in the
FAST user manual, the simple induction generator model is too simple to be used for a turbine
start up. Therefore the Thevenin equivalent for a three-phase induction generator was used. The
parameters for this model were taken from a generator with similar characteristics. In the case of
the normal operation, the simple induction generator which describes a linear relation between
torque and speed was used. The parameters from this simple model were obtained from
manufacturer specifications.
58
Generator torque
Generator torque
Gen
ener
era
ator speed Gener
era
ator speed
Figure 52. FAST_AD generator models [18]
The emergency braking maneuver consists of a sequence of events in which the generator goes
offlineeliminating the applied torqueand the emergency brake is deployed after a short
period. An emergency braking maneuver creates one of the harshest loads that the wind turbine
system can undergo, due to the high torsional loads and continuous reversals of the loading.
Figure 53 is a graphical representation of a braking maneuver, simulated with FAST for the
turbine modeled in this report. The event is performed at rated wind speed under high turbulence.
The figure also shows the different stages of the braking event.
Brake fully
stops the shaft
Normal operation under
turbulent condition
Vibrations remain
on the system
Loss of grid
Brake is engaged
59
The loading conditions were used as an input for the Stage 4 model. This model was chosen
because it has the greater number of degrees of freedom and the validity of its output is more
tangible. The input load chosen was a braking event from a constant-wind condition at a rated
wind speed. The constant wind speed was chosen because of its numerical simplicity; turbulent
input files tend to be numerically heavier. The last section of the braking event was chosen due
to the fully reversed loading conditions. The aim is to show possible output files that could be
used for further study, but which still capture the dynamical scenario.
Figure 54 shows the angular position change of the planet carrier. This is important because
changes in the angle of the planet carrier result in poor tooth-load distribution of the gears.
Uneven load distribution on gear teeth can result in micropitting, which would progress into
macropitting and is a common failure mechanism of gearboxes in the industry. The angular
misalignment of the planet carrier also results in uneven load distribution between the planets
and the ring. The same happens for the sun-planet interaction if the configuration of the gearbox
does not include a floating sun.
Figure 55 depicts axial displacements in the intermediate stage shaft. These deflections are
driven by the axial loading generated from the helix angle of the gears. The force generated by
the axial loading is a possible reason for bearing failures. Notice the step midway through the
oscillation; this is consequence of the effects of backlash. The data collected for this figure was
gathered at the final stages of the oscillation so that the influences of the backlash recorded. The
backlash interaction is much less noticeable with larger oscillations. It can also be said that the
backlash contributes the nonlinear behavior of the stiffness.
60
Model Comparison
This section compares the different models under operational loading conditions. The input load
cases implemented were generated by FAST_AD, simulating operation under turbulent
conditions at rated wind speed and the simple induction generator model. Models presented here
focus only on the gearbox, therefore the interaction of other components such as generator and
rotor are not considered. This leads to the implementation of input parameters such as angular
velocity and torque to be used in the models. To obtain the dynamical interactions of the internal
components of the gearbox, the angular velocities of the rotor were used as input on the models
low-speed shaft. Loads inherent to the high-speed shaft from FAST_AD were applied as an
opposing torsional load to the high-speed shaft of the gearbox, which revealed the internal load
interaction of the gearbox.
The first parameter compared between the different drivetrain models was the torque at the high-
speed shaft (which was the input torque), and the torque at the low-speed shaft (which was the
output torque). Figure 56 gives more insight on the previously noted loading condition.
61
Applied angular
velocity
High-speed shaft
Gearbox
Low-speed shaft Applied
torsional loading
Figure 56. Model comparison loading configuration
The Stage 1 model is the simplest model, and is far too simple to be compared to the other
models. It does not have a gear ratio, therefore the magnitude of the output is the same as the
input. A simple solution is to adjust the generator by increasing its inertia; this gives a better
approximation of the dynamical behavior, although it is still too rudimentary to mimic the true
behavior of the gearbox. Additionally, the minute changes in torque cannot be captured by using
a single large spring to represent the entire drivetrain. Figure 57 shows the torque relations
described above. Notice that the magnitudes are quite similar but have opposing signs.
Figure 57. The input torque and output torque for the Stage 1 model
The same approach was used for the rest of the models, although a much closer response of the
torsional behavior of the system was obtained. Figure 58 depicts the input torque and output
torque of the gearbox. Due to the increment in torque gained from the gear ratio, two
independent scales had to be used. Note that the response obtained more closely follows the
characteristics of the input torque. Due to the similarity in the response of all models, however,
only the Stage 3 model is shown.
62
Figure 58. Input torque and output torque comparison for Stage 3
A comparison between the torque outputs of all models also was performed, and is shown in
Figure 59. Note that the responses of the models are identical, and that torque for each is
superimposed.
The intermediate shaft of the gearbox was chosen as a point of comparison between the stages
because it is between the boundary conditions and it better captures the differences in the
complexity of each stage. The parameter compared was the angular velocity of the intermediate
shaft as taken at the gear. Figure 60 shows the recorded angular velocities of the simulation. The
response obtained generally follows the same form, although it is noticeable that Stage 2which
is the stage of least complexity in this comparisonhas higher excitation than was expected at
certain points of the simulation.
63
A closer examination of the same response reveals better insight into the importance of gear-
tooth interaction. Note that the response obtained from Stage 2 is less complex than the
responses from Stage 3 and Stage 4. The responses from Stage 3 and Stage 4 closely follow each
other, although slight differences exist due to the additional degrees of freedom in Stage 4.
Figure 61 compares the joint forces seen in the different stages. Stage 2 is not meant to include
the loading of bearing, therefore its response is nearly zero. The results for Stage 3 and Stage 4
are quite similar, regardless of the single degree of freedom joint used in Stage 3.
64
Based on the previous comparisons, it seems that the model of Stage 1 is a good approach to
perform a preliminary determination of stiffness that can be used to validate models of more
complexity. This model, however, does not represent the actual behavior of the gearbox closely
enough to be used for a more detailed analysis.
For Stage 2, the inclusion of all the components of the gearbox gives a much closer
representation of its true behavior. This stage has an extremely close torsional response in
comparison to the more complex models. Comparison of the angular velocity reveals that the
behavior of the Stage 2 model follows the general behavior expressed by models of higher
complexity. It already is noticeable, however, that the influences of gear-tooth contact are not
accounted for in this model.
For Stage 3, the implementation of the gear-tooth elements provides a good representation of the
behavior of the gearbox; depicting with better detail than Stage 2 the vibrational behavior and
various excitations generated by the gear-tooth contact. This model has only one DOF per
rotational body, therefore its computational time and stability are much better as compared to
Stage 4. The implementation of only one degree of freedom per rotational body, however, results
in the joined forces of the entire shaft being expressed as a single point. This quantity does not
reveal the loading distribution among multiple bearings on a shaft. In the example presented the
concurrence of both curves is very good, but with other configurations and the implementation of
bearing clearances this situation might change. An important addition to the loads revealed by
the Stage 3 model is the axial loading generated as a result of the helix angle in the gears. The
comment made regarding load distribution among bearings also applies to this type of loading
condition.
Stage 4 is the most complex model and therefore reveals the most information about the system.
Although for many parameters it follows the response of Stage 3 quite closely, it additionally
expands and reveals important information for the loading conditions of the bearings and gears.
The most obvious of these additions is angular misalignment of shafts and other components.
This stage shows the loading distribution among the different bearings of the gearbox, providing
valuable insight into the loading conditions of particular bearings.
65
Knowledge of the true loading conditions of the bearings, as well as the gears, potentially can
enhance the understanding of the unexplained bearing failures currently seen in the industry.
Additionally, models such as Stage 2 could be of great use for the design of other components of
the drivetrain. Nevertheless, models such as Stage 4 should be implemented for the collaboration
and iterative process between the gearbox manufacturer and the bearing supplier.
Pre-Design Process
Just as for any other industrial manufacturing process, the economical feasibility of a project is
evaluated to see if the project can flourish. This typically is done using a cost of energy study.
This study relies on a number of studies of existing machines that rate turbine size, head mass,
and drivetrain configuration to their economical viability. To calculate the cash flow return, this
study also considers the site where the turbines will be installed [26]. Cost of energy studies have
revealed that there is a considerable range of viable turbines sizes; however, the construction of
wind farms near populated areas has created the demand for quieter machines. Over time, this
has driven the industry to design and build larger machines.
From this original configuration the gear design is carried through. This includes calculation of
number of gear teeth and module and the center distances. Next come rating calculations,
including the allowable surface and bending stresses, and safety factors. Shaft dimensions also
are defined during this stage, which allows for the initial bearing selection that usually is
obtained from a sizing catalogue. To corroborate the initial selection, this is followed by the
cooperation of the bearing manufacturers that use in-house rating capabilities.
At this point a life analysis is performed on the selected gears. These fatigue calculations
typically are based on the Miner Palmgren method, which calculates the cumulative fatigue
damage due to the variable load spectrum. Common practice is to maintain constant ratios
66
between pitch diameter and face width. If the desired life is not met, then dimensions are
increased proportionally to meet the desired life requirements. This is closely followed by the
introduction of micro-geometry to reduce local loading peaks due to elastic deformation of the
bearings, shafts, and gear bodies.
Next the interfaces are satisfied. These typically include connections to the main shaft, as well as
the high-speed shaft, and design and interaction of the torque arms with the bedplate. Several
iterations could be required to satisfy the design objectives.
Aspects of the manufacturing process also must be discussed as part of the iterative process. This
includes the final refinements of the design, material characteristics, and heat treating, in addition
to outsourcing requirements for casting and forging. Final details such as oil cooling, sensors and
data interfaces, corrosion protection, and noise also are considered.
The final part of the design process is the manufacturing of two prototype gearboxes. These are
used for testing under specified loading conditions using the in-house dynamometer. The tested
elements are dissected and the wear is assessed to identify and fine-tune any changes that must
be made to the design before it is released for limited series 0 production and field testing in
operating turbines at various test sites. After the completion of field testing, any modifications
required by field-test results are implemented. This is followed by the release of the design for
large-scale production and sale.
Gear Design
The gears and internal components of the gearbox are designed simultaneously with the load-
path analysis. This process follows the AGMA 6006-A03 standard which dictates guidelines
directly tailored for the design of wind turbine gearboxes. The standard specifies the use of the
full-load spectrum and Miners rule cumulative fatigue method, and mandates separate safety
factors for pitting and bending.
67
Vibration Analysis
A Campbell diagram should be created using the calculated masses and stiffness. This design
predicts vibration and resonances inherent to the turbine. Mass and stiffness might have to be
adjusted to keep resonances out of the operating range.
Controls
Based upon inertias, pitch, and yaw rates, alternative control systems can be integrated into the
simulation. The controls will have an influence on the load cases and on the interaction of the
overall system to extreme loads. The new loads will be put through the simulation process again,
and perhaps be fine-tuned.
Estimated Masses
design Stiffness
Gear ratios
Aero elastic simulation
(Loads generation)
Feedback to
regenerate load cases Size components
according to
calculated loads
Gear design
Load path
calculation Preliminary
bearing selection
Bearing selection
Noise reduction Refine design
Micro geometry
Vibrations analysis
Controls algorithm
68
NonVertically Integrated Design Process
Develop Rotor and System Loads
This section develops the rotor of the turbine in great detail (see also Figure 63). The remaining
components of the turbine are estimated from previous designs. This information is sufficient for
calculating (by means of aeroelastic simulation code) the first draft for a complete loads
document.
Drivetrain Definition
At this point a definition of the drivetrain requirements is developed. This includes the gear ratio
of the drivetrain as well as the requirements and specifications of the interfaces between the
different components of the drivetrain. The reaction forces between interfaces also are defined.
Drivetrain Specification
Next, both previous steps are combined to create a document with the specification required for
the gearbox manufacture. This document includes a load matrix describing time series for
torsional and nontorsional loads and extreme load cases, and interface constraints and design
parameters such as safety factor requirements and stress curve selection.
Design Selection
A candidates designs undergo the iterative process of satisfying the requirements for cumulative
fatigue and the appropriate safety factors are applied. This is followed closely by a prototype
design review in which the necessary details to fabricate a prototype are discussed and
determined.
69
Estimated Masses
design Stiffness
Gear ratios
Aero elastic simulation
(Complete loads document)
Drivetrain
definition
Gearbox Bearing
provider provider
Prototype and testing
Series 0 testing
Figure 63 Nonvertically integrated design process
As shown in Figure 60, the process is quite horizontal and does not include a great deal of
iteration other than that of the independent subsystems. The overall calculations and ratings
therefore are based on the initial assumptions for masses and stiffness, which define the load
document. The reason for this situation stems from the interaction between the wind turbine
manufacturers and the gearbox vendors involved in the process, rather than the definition or
layout of the process itself.
70
To be able to update the load document and integrate the iterative process into this design
scheme, a great deal of information must be shared so that the wind turbine designer can create a
comprehensive dynamic model. This design process involves a number of competing gearbox
vendors, therefore the wind turbine manufacturer willingly shares only limited information. The
gearbox vendors most likely also work for other wind turbine manufacturers, thus information
from any particular wind turbine design is not shared. These informational barriers are the
disengaging factors of the feedback process.
The simple model used by the aeroelastic simulation codes commonly resembles the model in
Stage 1, therefore the entire drivetrain is represented by a single degree of freedom. The
advantage of this model is the simplicity of its input parameters, and that a minimal amount of
data is given to the wind turbine manufacturer.
Stage 3 requires a greater level of trust between parties. It requires providing the overall
configuration of the gearbox, as well as the number of teeth and module. If the tooth geometry is
sensitive information, then the geometry ratios described by the American Gear Manufacturers
Association (AGMA) standard could be used. Regardless of the tooth geometry, the tooth micro
geometrywhich currently is a significant area of research for noise reduction and efficiency
is not shared.
Stage 4 involves the parameters noted for Stage 3 with the addition of the required bearing
stiffnesses. Although the bearing stiffnesses do not have a great influence on the torsional
behavior of the system, they are quite important for vibration analysis. In this case generalized
stiffness data from the manufacturers are required. The stiffness response can be provided
without revealing any of the geometrical characteristics of the bearing. The analysis methods
used by the bearing manufacturer to calculate this stiffness are not revealed, which can protect
the data integrity. Additionally, the present study did not seek to determine the sensitivity of a
particular bearing stiffness, and a general stiffness for all bearings could be used thereby
reducing the amount of data shared.
71
The advantage of the progressive model approach is that it enables not only increasing complexity
among models but also the possibility of a complexity reduction. The response of a more
complicated model can be mimicked through implementation of complex stiffness parameters for
a simpler model. For example, the Stage 3 model can be reduced to the Stage 2 model by
implementing more complicated stiffness responses. Nonlinearities can be implemented to the
stiffness and damping coefficients, and prescribed excitation can account for tooth interaction
vibrations. This creates a model with a response that is closer to a more complicated model, even
though all the sensitive parameters are hidden under the nonlinear stiffness. It is evident that the
response of a system with fewer degrees of freedom would not be as accurate and would not
mimic all the desired parameters. Nevertheless, using this approach allows creation of models that
are closer to reality without necessitating the compromises generally involved; consequently, the
loading parameters can be redefined to a more accurate level.
The detailed multibody system simulation (MBS) truly captures the dynamical behavior of all
moving components. The approach, however, does not predict the minute mechanical and
structural behavior, which necessitates the cooperation between different disciplines. For
example, the creation of load cases for particular components of the drivetrain can be created
from the MBS simulation. These load cases can be used in the finite element model, which can
reveal in greater detail the mechanical behavior of the internal components such as tooth-load
distribution and tooth-bending characteristics. The load conditions cover a wide spectrum of
loading and load variations, which are narrowed to the load conditions of most relevance.
The premise of the research discussed in this report was to create a baseline model for a wind
turbine drivetrain. Models were developed with increasing complexity and correlated in a
progressive manner; they also were validated with experimental data collected from a
representative wind turbine in the field. The progressive models created are representative of the
gearbox and drivetrain standard of the industry, and include turbines that have output ranging
from 750 kW to 5 MW.
Using FAST-AD software, an aeroelastic model also was created for the same representative
field turbine. The aeroelastic model was used to create realistic load cases that could be input
into the multibody system dynamic models. The results from the multibody system simulation
from the input load cases were used to create load cases of the internal components of the
gearbox that could be used for further, more-detailed analysis, which could be better integrated
into the design process.
Different branches of the design process for the drivetrain of wind turbines also are discussed,
including the fully integrated design and the coupled design. This report also examines the
challenges in the data sharing between the wind turbine designers and the gearbox providers with
72
respect to the design process. These barriers to information accessibility produced by the fierce
competition and inherent proprietary data prevent the iterative nature of the design process.
This report suggests using the progressive multibody system dynamic approach as a data-sharing
sanitization tool. Using a minimal amount of sensitive data, comprehensive and detailed models
can be created to enhance the load-prediction methods of the design process, and return its
iterative characteristics to the non-vertical design process.
The validation of the models using the obtained experimental data provides great insight into the
correlation between theoretically calculated values and experimental empirical response. The
theoretical data used in the various models had a disparity that was not greater than 10% for the
stiffness coefficients and was not greater than 20% for the dampening coefficients. This finding
is valuable because the theoretical values can be calculated using known physical dimensions of
the gearbox; the response obtained will not differ greatly relative to the true response.
Comparison of the different models revealed that models having the level of complexity of Stage
1 should not be used for more than a rough estimation of the dynamical behavior. Stage 2
through Stage 4 gave concurrent responses for the torsional behavior of the different components
of the system. In the case of angular velocities and acceleration, the models with greater
complexity revealed a more complex loading response generated by the gear-tooth interactions.
The Stage 4 model proved to be more valuable for the bearing- and gear-failure predictions,
because it is capable of revealing misalignments of the different components of the gearbox.
It is worthwhile to note that the parameters used to validate the simulation were strictly linear.
Thus, the simulated response does not exactly match the response obtained from the
experimental data. The change in stiffness can be thought of as a change in frequency of the
obtained response. The change in frequency is a consequence of clearances and component
interactions that become more prominent as the magnitude of the excitation is reduced. Figure 64
shows the normalized change in the stiffness coefficient collected from the experimental data.
Models such as those for Stage 1 and Stage 2 do not depict the nonlinear behavior. In models
such as those for Stage 3 and Stage 4, the effect of backlash has an influence on the overall
stiffness of the system. This influence results in a change of frequency or stiffness that is similar
to the change seen in the experimental response, although the correlation is not as strong.
Additionally, the neglect of all the nontorsional flexibilities of the gearbox such as the shaft and
housing, as well as the flexibilities of the bedplate, also could have influenced the misalignment
and loading conditions of the drivetrain components.
73
Future Work
Further instrumentation and data collection will follow to better validate the current models. The
data collected will include dynamometer as well as field data. The parameters collected from the
dynamometer will include internal displacements, misalignments, and vibrations. The field data
will include real operational inputs such as torque, and angular velocities of input and output.
This information can be used to fine-tune the existing validated models, and to assure that the
response simulated correlates with the true behavior of the machine.
The models currently built employ arbitrary stiffness and damping coefficients for the bearings.
This approach reiterates the importance and capabilities of the MBS approach, although it does
not reproduce the true behavior of the system. Further research revealing the nonlinear
characteristics of the stiffness and damping coefficients of the bearings would be valuable.
Research correlating bearing failure modes to specific load conditions also would be of great use,
because particular loading conditions could be identified under simulated operational conditions.
The shafts in the Stage 4 model currently represented by rigid bodies will be substituted by
flexible beam elements. These beam elements can handle nonlinearities of the system, and will
allow eccentricities and misalignment of the gears in a more accurate manner. The shaft
flexibilities can be obtained by a finite element analysis to find the modal behavior of the shaft
and the modal stiffness. Performing a Guyan reduction reduces the number of degrees of
freedom. This reduced element is imported into the multibody system simulation, enabling the
system to represent the dynamical behavior and flexibilities [8]. Nevertheless, the Guyan
reduction keeps the computation light, thus allowing for simulation of longer events. The same
modal approach can be implemented for gearbox housing, simulating the flexibilities of the
housing under dynamical loading. The deformation resulting from this flexibility also can
contribute to gear misalignment and internal compliances of the system.
74
Further analysis will follow the dynamical analysis, including finite element analysis of the
gearbox, and the implementation of the generated dynamical input load cases. This detailed
model will reveal important factors such as tooth load distribution, which can be correlated to
fatigue and other failure modes. To truly capture the overall dynamical coupled behavior of the
system the dynamical model should be extended to include other components of the drivetrain,
such as generator and rotor.
75
References
1. Wind Turbine Data Summary Table 2. Wind Stats Newsletter. Volume 16 19, Numbers 1 4,
2003 2006. Volume 20, Number 1, 2007.
2. Walford, C. (2006). O&M Cost Model Quantifying the Influrnce of Reliability, Wind Turbine
Reliability Workshop. Global Energy Concepts Seattle, WA.
cwalford@globalenergyconcepts.com
http://www.sandia.gov/wind/2006reliability/wednesday/09-chriswalford%20.pdf
3. Gasch, R.; Twele, J. (2002). Wind Power Plants Fundamentals: Design, Construction and
Operation. Berlin: Solarpraxis AG.
4. Poore R.; Lettenmaier, T. (2002). Alternative Design Study Report: WindPAC. Kirkland
Washington: Advanced Wind Turbine Drivetrain Design Study Global Energy Concepts,
LLC.
5. Manwell, J. F.; Mc Gowan, J. G.; Rogers, A. L. (2002). Wind Energy Explained: Theory,
Design and Application. Chichester, NY: Wiley.
6. Norton, R. L. (2006). Machine Design: An Integrated Approach, 3rd edition. Upper Saddle
River, NJ: Pearson Prentice Hall.
7. Ackerman, T. (2005). Wind Power in Power Systems. Hoboken, NJ: John Wiley.
8. Drago, R. J. (1988). Fundamentals of Gear Design. Boston, MA: Butterworths.
9. Spotts, M. F. (1951). Design of Machine Elements, 4th edition. London: G. Allen & Unwin.
10. ANSI/AGMA (1980). AGMA Standard 110.04. Nomenclature of Tooth Failure Modes
(reaffirmed 1989).
11. Jonkman, J. M. (2003). Modeling of the UAE Wind Turbine for Refinement of FAST_AD.
NREL TP-500-34755. Golden, CO: National Renewable Energy Laboratory.
12. Popov, E. P.; Balan, T. A. (1998). Engineering Mechanics of Solids, 2nd edition. Upper
Saddle River, NJ: Prentice Hall.
13. Angeles, J. (ed.) (1995). Kinematics and Dynamics of Multi-Body Systems. Wien, NY:
McGill University and A. Kecskemethy Gehard Mercator University.
14. INTEC GmbH. SIMPACK Basics Training 1. Wessling, Germany.
15. INTEC GmbH. SIMPACK User Manual. Force Element Library. Wessling, Germany.
16. Ramamurti, V. (2000). Mechanical Vibration Practice with Basic Theory. Boca Raton, FL:
CRC Press/Narosa Publishing House.
17. Palm, W. J. III (2007). Mechanical Vibration. Hoboken, NJ: John Wiley.
18. Jonkman, J. NWTC Design Codes, FAST, An Aeroelastic Design Code for Horizontal Axis
Wind Turbines. http://wind.nrel.gov/designcodes/simulators/fast (accessed Sept. 18,
2008).
19. Peeters, J. L. M.; Vandepitte, D.; and Sas, P. (2005). Analysis of Internal Drive Train
Dynamics in a Wind Turbine. Wind Energy, 9(12), pp. 141161.
20. Rivin, E. I. (1999). Stiffness and Damping in Mechanical Design. New York: Marcel
Dekker.
21. INTEC GmbH. SIMPACK User Documentation. III-FE:14 Gearbox with Elastic
Transmission. Wessling, Germany.
22. INTEC GmbH. SIMPACK User Documentation. III-FE:225 Gear Pair. Wessling, Germany.
23. INTEC GmbH. SIMPACK FEMS Training Course. Wessling, Germany.
24. INTEC GmbH. SIMPACK Basics Training Course. Wessling, Germany.
76
25. Gipe, P. (1993). Wind Power: Renewable Energy for Home, Farm, and Business. Post Mills,
VT: Chelsea Green Publishing Co.
26. National Wind Technology Center (2006). Baseline Cost of Energy.
http://www.nrel.gov/wind/coe.html (accessed Sept. 18, 2008).
27. NEG Micon A/S (2000). Weights, Dimensions and Transport Guidelines NM 750/48.
TIC247002 GB, Provided by Turbine operator Ken Bolin Excel Energy.
28. Gold, P. W.; Schelenz, R.; Frenschek, W.; Klein, A.; Moller, D. (2004). Simulation of the
Three-Dimensional Vibration Behavior of a Wind Energy Plant. SIMPACK User
Meeting, 2004. Available at http://www.simpack.com/downloads/pdf/
um04_rwth_moeller.pdf (accessed Sept. 18, 2008).
29. Moriarty, P. J.; Hansen A. C. (2005). AeroDyn Theory Manual. NREL/TP-500536881.
Golden, CO: National Renewable Energy Laboratory.
30. Jonkman, J. M. (2003). Modeling of the UAE Wind Turbine for Refinement of FAST_AD.
NREL/TP-500-34755. Golden, CO: National Renewable Energy Laboratory.
31. ANSI/AGMA. (2004). AGMA Standard 2001-D04. Fundamental Rating Factors and
Calculation Methods for Involute Spur and Helical Gear Teeth.
32. Heege, A. (2003). Computation of Dynamic Loads in Wind Turbine Power Trains. DEWI
Magazine 23, pp. 5964. Available at http://www.dewi.de/dewi/fileadmin/pdf/
publications/Magazin_23/08.pdf (accessed Sept. 18, 2008).
33. Wilcoxon Research. Bearing Failures: Causes and Cures. http://www.wilcoxon.com/
knowdesk/bearing.pdf (accessed Sept. 18, 2008).
34. DeLange, G. Failure Analysis for Gearing. http://www.elecon.com/gearworld/dat-gw
failure.html (accessed Sept. 18, 2008).
35. Schaeffler Group Industry (2006). International Standard ISO 1524. Ad-hoc Meeting
Orlando, Florida.
36. Geartech (1999). Gear Failure Atlas. Townsend, MT: Geartech.
37. GEEA. Nordex diagram. http://www.geea.org/IMG/jpg/vue_eolienne_nordex1000.jpg
(accessed Sept. 18, 2008).
38. Mauer, L. (2006). The New, Powerful Gearwheel Module. SIMPACK User Meeting 2006
INTEC GmbH, Wessling, Germany. Available at http://www.simpack.com/downloads/
pdf/um06_intec_mauer.pdf (accessed Sept. 18, 2008).
77
The thrust of the turbine can be calculated using the principle of linear momentum, because the
thrust is equal in magnitude and opposite in direction to the change in momentum [5]. As the
control volume increases in area and the outgoing velocity reduces, the mass flow remains
constant. Thus, the change in angular momentum is solely dependent on the change in flow
velocity. The thrust can be calculated from the following expression.
Trot = m (V1 V3 )
If there is no reduction in the flow velocity and the flow passes without resistance, then there is
no change in momentum and no thrust or power is generated. Furthermore, if the flow is stopped
completely then the mass flow rate is zero; consequently, the change in momentum is zero as
well. This results in an optimum velocity ratio of 1:3 between the outgoing velocity (V3) at the
end of the control volume and the ingoing velocity (V1) entering the control volume. This allows
the calculation of the power coefficient, as well as giving a max power of Cp = 0.59, which is the
power coefficient for an ideal turbine. [3]
78
that for every action there is an equal and opposite reaction. Consequently, an angular
momentum that rotates in a direction opposing the rotor is induced on the wake.
As was done in the previous analysis, the principle of conservation of momentum is used to
calculate the corresponding load. The main difference is that the control volume is subdivided
into a number of annuli to better account for the change in rotation of the wake. In this case, the
expression for torque experienced on the rotor can be derived using conservation of angular
momentum.
To better understand the simulation theory, two important quantities should be defined. The first
is the lift force, which is the driving force of any lift devise. Lift force can be defined as the
mechanical force generated as a solid travels through a fluid; it is perpendicular to the inflow of
the solid. More specifically, lift is generated as a fluid travels over the airfoil by creating lower
pressure above the profile. This occurs due to the convex surface of the profile increasing the
flow velocity and inherently reducing the pressure. The concave side or bottom side, in contrast,
generates greater pressure. These low and high pressure differences result in lift forcethe low
pressure pulls the profile and the high pressure pushes the profile [30]. The other important
quantity is the drag force. Drag force is a force which opposes the motion of the airfoil as it
travels through the fluid, and most applications seek to minimize this force. Drag force is
generated by the viscous effects of fluid traveling over the profile and by unequal pressure
distribution on the airfoil [30]. These two measures result in a friction force that is parallel to the
inflow [5].
Lift and drag can be characterized as dimensionless parameters called lift and drag coefficients.
The values of these coefficients greatly depend on the airfoil profile and angle of attack, and are
less dependent on the size and the relative speed of the flow. Coefficient of lift, Cl, and
coefficient of drag, Cd, can be calculated using the following expressions. [11]
79
L / ls D / ls
Cl = Cd =
1 1
U 2 c U 2 c
2 2
Here, L is the lift force, D is the drag force, and ls is the airfoil span. In the denominator, the
dynamic forces per unit length are taken into account; is the fluid density, U is the velocity of
the undisturbed fluid flow, and c is the cord length of the airfoil. [5]
These parameters generally are determined experimentally in a wind tunnel [3], where scaled-
down airfoils are tested under controlled flow conditions. The lift and drag forces are recorded at
a range of angles of attack and the coefficients are calculated (see Figure 67).
At lesser angles of attack the lift coefficient increases linearly, with the angle of attack increasing
to approximately 10. The curve then tends to flatten before reaching its maximum value [3].
This is the result of the initial separation of the flow above the airfoil. As the angle of attack
increases the separation becomes more evident, and a drastic reduction of the lift and increase in
the drag can be seen. At greater angles of attack the behavior of the airfoil resembles a flat plate.
Wind turbines operate in this range, therefore the behavior is estimated using this assumption for
the entire 360 range [5].
The blade element theory subdivides the blade along its length into a number of elements. These
elements sweep a number of annuli as the rotor rotates. The theory assumes that there is no
interaction between each blade element, and that the behavior of the blade profile follows the
two-dimensional lift and drag characteristics of the profile [11]. The lift and drag forces therefore
can be calculated with the previously described lift and drag coefficients.
80
The missing quantity is the angle of attack for each individual section of the blade. The angle of
attack is dependent on the inflow angle and the particular angle of twist of the blade. The inflow
angle is the result of a vectorial addition involving the tangential velocity of the blade element
and the incoming wind. The tangential velocity of the particular section of blade can be
calculated from the angular velocity of the rotor and the radial distance of the element with
respect to the axis of rotation. The calculated lift and drag forces must be decomposed into
perpendicular and parallel components with respect to the rotor sweeping plane. This
decomposition results in the thrust for the perpendicular force and the torque for the tangential
force. This procedure must be repeated for each individual element along the blade, and the
overall thrust and torque of the rotor are the sum of all the element forces composing the
rotor [29].
81
Form Approved
REPORT DOCUMENTATION PAGE OMB No. 0704-0188
The public reporting burden for this collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources,
gathering and maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this
collection of information, including suggestions for reducing the burden, to Department of Defense, Executive Services and Communications Directorate (0704-0188). Respondents
should be aware that notwithstanding any other provision of law, no person shall be subject to any penalty for failing to comply with a collection of information if it does not display a
currently valid OMB control number.
PLEASE DO NOT RETURN YOUR FORM TO THE ABOVE ORGANIZATION.
1. REPORT DATE (DD-MM-YYYY) 2. REPORT TYPE 3. DATES COVERED (From - To)
February 2009 Technical report
4. TITLE AND SUBTITLE 5a. CONTRACT NUMBER
Gearbox Modeling and Load Simulation of a Baseline 750-kW Wind DE-AC36-08-GO28308
Turbine Using State-of-the-Art Simulation Codes
5b. GRANT NUMBER
11. SPONSORING/MONITORING
AGENCY REPORT NUMBER
16. SECURITY CLASSIFICATION OF: 17. LIMITATION 18. NUMBER 19a. NAME OF RESPONSIBLE PERSON
OF ABSTRACT OF PAGES
a. REPORT b. ABSTRACT c. THIS PAGE
Unclassified Unclassified Unclassified UL
19b. TELEPHONE NUMBER (Include area code)
F1147-E(10/2008)